WO2009125349A2

WO2009125349A2 - Multi-sensor apparatus and method for monitoring of circulatory parameters

Info

Publication number: WO2009125349A2
Application number: PCT/IB2009/051469
Authority: WO
Inventors: Ehud Baron; Amir Haiman
Original assignee: Cardiosigns Ltd.
Priority date: 2008-04-10
Filing date: 2009-04-07
Publication date: 2009-10-15
Also published as: WO2009125349A3

Abstract

A system for measuring blood circulation parameters in a patient, comprising: a) a pressing element; b) a contact microphone which is acoustically coupled to the patients body by the pressing element at a location adjacent to an artery, the microphone thereby producing a data signal indicative of blood circulation in the artery; and c) a controller which integrates in time the data signal from the microphone, thereby determining a blood pressure as a function of time, at least up to an unknown integration constant.

Description

MULTI-SENSOR APPARATUS AND METHOD FOR MONITORING OF CIRCULATORY PARAMETERS

RELATED APPLICATION

The present application claims benefit under 35 USC 119(e) from US provisional patent application 61/071,052, filed on April 10, 2008.

The contents of the above document are incorporated by reference as if fully set forth herein.

FIELD AND BACKGROUND OF THE INVENTION

The present invention, in some embodiments thereof, relates to a method and apparatus for measuring circulatory parameters and, more particularly, but not exclusively, to a method and apparatus for beat-to-beat monitoring of blood pressure and other circulatory parameters.

Known methods of measuring blood pressure include the following: 1) The tonometric method of measuring pulse wave shape uses a pressure sensor or a displacement sensor (described in US patents 4,960,128 to Gordon and 4,669,485 to Russell), placed on an artery. 2) The oscillometric method determines systolic and diastolic pressure by exerting a slowly changing pressure, manually or automatically, on an artery, and observing the pulse with a pressure sensor. This is described, for example, in US patent 6,241,679 to Curran, and in US patent 6,432,060 to Amano.

3) The auscultatory method involves exerting enough pressure on an artery, for example with a cuff, to occlude blood flow completely, then slowly decreasing the pressure and detecting Korotkoff sounds distal to the cuff, using a stethoscope, or automatically using a microphone and a computer as described by US patent 4,459,991 to Hatschek.

4) Pulse wave velocity is measured, and blood pressure is inferred from that, often using the oscillometric or auscultatory method to calibrate the blood pressure.

Pulse wave velocity can be found using any of a variety of sensors to measure the difference in pulse wave timing at two different distances from the heart. For example one sensor is near the heart, and the other sensor is on an arm, a leg, or an ear lobe. The sensors can include pressure sensors (described in US published patent application

2003/0199770 to Chen et al), microphones (described in US patent 5,237,997 to Greubel et al), photople thy sinography (PPG) sensors (described by Greubel et al), pulse oximetery sensors (described in EP 1,344,489 to Friedman et al), electric impedance plethysomography (IPG) sensors (described by Friedman et al, as well as in US patent 5,309,916 to Hatschek), laser Doppler sensors (described by US 5,309,916), ultrasound Doppler sensors (described by US 5,309,916, as well as in US patent 5,241,964 to McQuilkin), ultrasound imaging sensors (described by US 5,309,916), and electrocardiograph (ECG) devices (described by Greubel et al, as well as in US 2005/0663259 to Orbach).

Some of these sensors can also be used to measure pulse wave shape optically at a single location, for example pulse oximetry with red light as described in US patent 5,111,817 to Clark et al, and PPG with green and infrared light as described in WO 2007/097702 to Lindberg et al. US 2006/074322 to Nitzan describes using PPG to measure systolic blood pressure.

Pulse wave shape and pulse wave velocity can also be used to find other circulatory parameters of interest, such as arterial compliance and cardiac stroke volume. For example, US patent 6,955,649 to Narimatsu et al describes using the ratio of peak reflected wave pressure to peak incident wave pressure to evaluate atherosclerosis. Friedman et al describes using pulse wave transit time and blood volume measurements to detect changes in arterial compliance.

Japanese published patent application JP2000/316823 describes a method of measuring pressure wave velocity by using an actuator to induce a pressure wave in an artery, at a frequency much higher than the pulse frequency, and measuring the phase difference with a sensor located a few centimeters away along the same artery.

Two or more sensors can be used to obtain measurements that are more robust to noise. For example, US patent 4,459,991 to Hatschek describes a microphone which records Korotkoff sounds, which are identified automatically by a computer. The computer uses data from a pressure sensor to tell when Korotkoff sounds are likely to be present, and only analyzes sounds from those time intervals. Japanese published patent application JP2000/033078 describes using the pulse wave velocity, the cardiac cycle, and the volume pulse wave ratio, to obtain a more reliable measure of blood pressure, reducing the need for frequent calibration. US patent 5,853,364 describes using both

PPG data and IPG data from a finger, to estimate blood pressure using a hemodynamic model.

A description of various blood circulation parameters, how they can be calculated from pulse wave shape and pulse wave velocity, and how they can be used clinically, may be found in R. L. Chatburn and M. D. Lough, Handbook of Respiratory

Care, Second Edition, Year Book Medical Publishing, 1990, and is posted online at www.ventworld.com/resources/convert/Hemo.swf.

The CAFE Investigators, for the Anglo-Scandinavian Cardiac Outcomes (ASCOT) Investigators, "Differential impact of blood pressure-lowering drugs on central aortic pressure and clinical outcomes: principal results of the Conduit Artery Function Evaluation (CAFE) study," Circulation 113: 1213-1225 (2006), and an accompanying editorial by Suzanne Oparil and Joseph L. Izzo, Jr., "Pulsology Rediscovered: Commentary on the Conduit Artery Function Evaluation (CAFE) Study," Circulation 113: 1162-1163 (2006), describes the clinical value of measuring central systolic pressure, which can be calculated from the pulse wave shape, as opposed to only measuring peripheral systolic pressure, and the value of using drugs which can lower central systolic pressure.

Additional background art includes a description of blood pulse wave monitoring posted online at www.mi-labs.co.jp/RD2-e.htm (downloaded on April 9, 2008), US patent 7,344,502, to Tanabe, and K. Takazawa et al, Hypertension Research 30(3), 219- 228 (2007).

SUMMARY OF THE INVENTION An aspect of some embodiments of the invention concerns methods and apparatus for measuring blood circulation parameters in a patient, including pulse wave shape, pulse wave velocity, and systolic pressure, that have improved accuracy, comfort to the patient, and convenience.

There is thus provided, in accordance with an exemplary embodiment of the invention, a system for measuring blood circulation parameters in a patient, comprising: a) a pressing element; b) a contact microphone which is acoustically coupled to the patient' s body by the pressing element at a location adjacent to an artery, the microphone thereby producing a data signal indicative of blood circulation in the artery; and c) a controller which integrates in time the data signal from the microphone, thereby determining a blood pressure as a function of time, at least up to an unknown integration constant.

Optionally, the system also includes a pressure sensor which is coupled to the patient's body at the same or a different location adjacent to the artery, thereby producing a data signal indicative of a blood circulation in the artery, wherein the controller uses the data signal from the pressure sensor to determine the integration constant.

Optionally, the controller combines the integrated data signal from the microphone and the data signal from the pressure sensor to find a most probable blood pressure as a function of time. Optionally, the microphone is capable of detecting, above a noise level, rates of change of pressure of at least 20 mm Hg per second, for frequencies at least as low as 1 Hz, in no more than 0.1 seconds.

Optionally, the pressing element acoustically couples the microphone to the patient's body at a location adjacent to the radial artery, the brachial artery, the posterior tibial artery, the femoral artery, or the carotid artery.

There is further provided, in accordance with an exemplary embodiment of the invention, a system for measuring a pulse wave velocity in an artery of a patient, comprising: a) a first and a second sensor of arterial pulse data, in a positioning element which couples the sensors to the body of the patient adjacent to an artery, spaced at a known separation distance of less than 10 cm along the artery, each sensor associated with an output channel conveying a data signal and with a comparator that provides an indication of a time at which the data signal had a specific value; and b) a controller which measures a pulse wave delay time between the first and second sensor, by comparing time indicated by the comparators of the first and second sensors. Optionally, the known separation distance is less than 5 cm.

Optionally, the known separation distance is less than 3 cm.

Optionally, the two sensors measure a same kind of data.

Optionally, the two sensors are substantially identical. In some embodiments of the invention, the two sensors comprise one or both of a microphone and a pressure sensor.

Optionally, the specific value corresponds to systolic time for both sensors, or corresponds to diastolic time for both sensors.

Optionally, the indication of time for each sensor comprises an identification of a clock cycle of the comparator for that sensor, and the clock cycle is shorter than 0.1 millisecond.

There is further provided, in accordance with an exemplary embodiment of the invention, a system for estimating systolic blood pressure of a patient, comprising: a) a pressure applicator which applies a controllable pressure on an artery of the patient; b) a sensor which measures an amplitude of a pulse beat in the artery when the pressure applicator is applying the controlled pressure; and c) a controller which: i) varies the applied pressure over a range of pressures low enough not to occlude blood flow in the artery completely, over a period covering a plurality of pulse beats, while receiving data from the sensor; ii) determines the pulse amplitude as a function of the applied pressure; and iii) estimates the systolic pressure by extrapolating the data to an applied pressure at which the amplitude of the pulse beat would go to zero. Optionally, the controller varies the applied pressure by increasing it to value at which the pulse amplitude is reduced by a factor of at least two from its value at the lowest applied pressure in the range.

Optionally, the factor is at least five.

In some embodiments of the invention, the lowest applied pressure in the range is less than an expected diastolic pressure of the patient, and the controller also estimates the diastolic pressure from the pulse amplitude as a function of applied pressure. Optionally, the sensor is a contact microphone.

Alternatively, the sensor is a pressure sensor. Optionally, the system includes a second sensor, and one sensor is a contact microphone while the other sensor is a pressure sensor. There is further provided, according to an exemplary embodiment of the invention, a method for determining a blood pressure as a function of time, at least up to an unknown integration constant, for a patient's artery, comprising integrating in time a data signal indicative of a pulse beat in the artery, obtained from a contact microphone acoustically coupled to the patient's body adjacent to the artery. Optionally, the data signal is integrated for at least one pulse cycle, thereby obtaining a pulse wave shape of the pressure as a function of time.

Optionally, the method includes acoustically coupling the microphone to the patient's body adjacent to the artery, and generating the data signal indicative of the pulse beat in the artery. There is further provided, according to an exemplary embodiment of the invention, a method for measuring a pulse wave velocity in a patient's artery, comprising: a) receiving from each of two comparators in microprocessors associated with two sensors coupled to the patient's body adjacent to the artery and separated by a known separation distance along the artery, an indication of a time when a data signal from that sensor had a specific value; b) finding a pulse wave delay time from a difference between the times indicated for the two sensors; and c) calculating the pulse wave velocity from the separation distance and the pulse wave delay time.

Optionally, the method includes coupling the two sensors to the patient's body adjacent to the artery and separated by the separation distance, and generating the data signal from each sensor.

There is further provided, according to an exemplary embodiment of the invention, a method of estimating systolic blood pressure of a patient, comprising: a) applying a pressure varying over a range to an artery of the patient, over a period of a plurality of pulse cycles and substantially constant during each pulse cycle, with the maximum pressure in the range partially occluding blood flow in the artery but not completely occluding the blood flow; b) generating data on a pulse amplitude in the artery while applying the varying pressure; c) determining the pulse amplitude as a function of the applied pressure; and d) extrapolating the data to a pressure at which the pulse amplitude would go to zero, thereby estimating the systolic pressure.

Optionally, applying the varying pressure comprises increasing the pressure from the minimum pressure in the range to the maximum pressure in the range. Optionally, the method comprises stopping the increasing of the varying pressure, responsive to the data on pulse amplitude.

In some embodiments of the invention, the method also includes: e) repeating (a) through (d) one or more times; f) recording the estimated systolic pressure each time (d) is repeated; and g) finding an average of the recorded estimated systolic pressures, to obtain a more accurate estimate of the systolic pressure.

There is further provided, according to an exemplary embodiment of the invention, a system for measuring a breathing parameter of a patient, the system comprising: a) a set of one or more sensors suitable for measuring beat-to-beat blood pressure in an artery of the patient; and b) a controller which uses data from the sensors to detect beat-to-beat changes in one or more blood pressure parameters indicative of the patient's breathing cycle, thereby measuring the breathing parameter.

Optionally, the controller uses the data to measure a breathing parameter indicative of a sleep disorder.

There is further provided, according to an exemplary embodiment of the invention, a method of measuring a breathing parameter in a patient, the method comprising: a) measuring a blood pressure parameter in the patient for an interval comprising a plurality of breathing cycles; b) analyzing changes in the blood pressure parameter from beat to beat over a breathing cycle timescale; and c) extracting the breathing parameter from the changes in the blood pressure parameter.

Optionally, the method also includes using the breathing parameter to diagnose a sleep disorder. There is further provided, according to an exemplary embodiment of the invention, a method of obtaining homogeneous segments of data on blood circulatory parameters, comprising: a) generating data of at least one circulatory parameter as a function of time; b) identifying as homogeneous segments one or more time intervals, within each of which the parameter is relatively homogeneous; and c) calculating a representative sample of the data for each homogeneous segment. Optionally, calculating a representative sample comprises choosing a representative subset of the data in the homogeneous segment.

Alternatively or additionally, calculating a representative sample comprises finding an average over the homogeneous segment.

Optionally, the method also comprises identifying scaling segments of data from scaling events.

Optionally, the method also comprises identifying as noise segments, one or more time intervals within which the data has atypical values, and excluding the noise segments from consideration as homogeneous segments.

There is further provided, according to an exemplary embodiment of the invention, a system for determining a set of one or more blood circulation parameters, comprising: a) one or more sensors for obtaining a plurality of data signals pertaining to the blood circulation parameters; and b) a controller which, for each parameter, calculates a probability distribution for the value of the parameter from each data signal, and calculates an overall probability distribution for the value of the parameter by combining the probability distributions for each data signal.

Optionally, the system also includes a pressure applying element for applying a variable known pressure to an artery, wherein the data signals comprise pulse data obtained from the artery while the variable known pressure is applied to the artery, and the blood circulation parameters comprise one or both of systolic pressure and diastolic pressure.

Additionally or alternatively, the sensors comprise a microphone and a pressure sensor, the data signals comprise a first signal from the microphone and a signal from the pressure sensor, and the blood pressure parameters comprise information on pulse wave shape.

Optionally, the data signals also comprise a second signal from the microphone, the first signal being filtered to relatively increase lower frequencies, and the second signal being filtered to relatively increase higher frequencies. There is further provided, according to an exemplary embodiment of the invention, a method for determining a set of one or more blood circulation parameters, comprising: a) obtaining a plurality of data signals pertaining to the blood circulation parameters; b) calculating, for each parameter, a probability distribution for the value of the parameter from each data signal; and c) calculating an overall probability distribution for the value of the parameter by combining the probability distributions for each data signal.

Optionally, the data signals comprise pulse data obtained from an artery while a variable known pressure is applied to the artery, and the blood circulation parameters comprise one or both of systolic pressure and diastolic pressure.

In some embodiments of the invention, calculating a probability distribution, for one or both of the systolic and diastolic pressure, comprises: a) calculating from the data signal a probability distribution of the time of occurrence of the category of pressure; b) finding the value of the variable known pressure as a function of time for a range of time of the distribution of the time of occurrence; c) obtaining a measure of stiffness of the artery; and d) using the measure of stiffness of the artery to correct the value of the variable known pressure as a function of time for the range of time, to obtain the probability distribution for the category of pressure. Optionally, obtaining the measure of stiffness of the artery comprises obtaining a measure of stiffness based on a measurement of pulse wave velocity.

Additionally or alternatively, obtaining the data signals comprises obtaining at least one signal from a microphone and a signal from a pressure sensor, and the blood pressure parameters comprise information on pulse wave shape.

Optionally, obtaining at least one signal from the microphone comprises obtaining a low pass filtered signal and a high pass filtered signal from the microphone.

There is further provided, in accordance with an exemplary embodiment of the invention, a blood circulation monitoring system adapted to be held around a part of a patient's body, the system comprising: a) an actuator or fluid-filled bag that exerts a controllable level of pressure locally on an artery in that part of the body, but does not exert pressure completely around that part of the body; b) a vibration sensor or microphone that presses against the artery to sense vibrations; and c) a processing unit that controls the level of pressure applied to the artery, and receives data from the vibration sensor or microphone.

Unless otherwise defined, all technical and/or scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the invention pertains. Although methods and materials similar or equivalent to those described herein can be used in the practice or testing of embodiments of the invention, exemplary methods and/or materials are described below. In case of conflict, the patent specification, including definitions, will control. In addition, the materials, methods, and examples are illustrative only and are not intended to be necessarily limiting.

Implementation of the method and/or system of embodiments of the invention can involve performing or completing selected tasks manually, automatically, or a combination thereof. Moreover, according to actual instrumentation and equipment of embodiments of the method and/or system of the invention, several selected tasks could be implemented by hardware, by software or by firmware or by a combination thereof using an operating system. For example, hardware for performing selected tasks according to embodiments of the invention could be implemented as a chip or a circuit. As software, selected tasks according to embodiments of the invention could be implemented as a plurality of software instructions being executed by a computer using any suitable operating system. In an exemplary embodiment of the invention, one or more tasks according to exemplary embodiments of method and/or system as described herein are performed by a data processor, such as a computing platform for executing a plurality of instructions. Optionally, the data processor includes a volatile memory for storing instructions and/or data and/or a non-volatile storage, for example, a magnetic hard-disk and/or removable media, for storing instructions and/or data. Optionally, a network connection is provided as well. A display and/or a user input device such as a keyboard or mouse are optionally provided as well.

BRIEF DESCRIPTION OF THE DRAWINGS Some embodiments of the invention are herein described, by way of example only, with reference to the accompanying drawings. With specific reference now to the drawings in detail, it is stressed that the particulars shown are by way of example and for purposes of illustrative discussion of embodiments of the invention. In this regard, the description taken with the drawings makes apparent to those skilled in the art how embodiments of the invention may be practiced. In the drawings:

FIG. 1 schematically shows a blood circulation monitoring system mounted on a patient's wrist, according to an exemplary embodiment of the invention;

FIG. 2 schematically shows a perspective view of a wrist assembly and band of the system shown in FIG. 1;

FIG. 3 schematically shows a cross-sectional view of a processing unit and wrist assembly of a blood circulation monitoring system similar to that shown in FIG. 1, using a motor-driven force applicator, according to an exemplary embodiment of the invention; FIG. 4 schematically shows a cross-sectional view of a processing unit and wrist assembly of a blood circulation monitoring system using an air-bag force applicator, according to an exemplary embodiment of the invention; FIG. 5 is a block diagram of hardware for the blood circulation monitoring system of FIG. 4;

FIG. 6 is a flowchart for a "scaling event" using an oscillometric and auscultatory method to measure systolic and diastolic pressure, according to an exemplary embodiment of the invention;

FIG. 7 is a schematic plot of data obtained using the method of FIG. 6;

FIG. 8 is a schematic plot of distributions of probability of the systolic point occurring at different pulse beats, using the method of FIG. 6;

FIG. 9 is a flowchart for a "mini-scaling event" using an oscillometric and auscultatory method to measure systolic and diastolic pressure, keeping the applied pressure below the systolic pressure for improved patient comfort, according to an exemplary embodiment of the invention;

FIG. 10 is a schematic plot of data obtained using the method of FIG. 9;

FIG. 11 is a flowchart of segmenting and compressing data from a blood circulation monitoring system, according to an exemplary embodiment of the invention;

FIG. 12 schematically shows exemplary displayed data generated by the method of FIG. 11;

FIG. 13 is a schematic plot of exemplary microphone and pressure sensor data from a blood circulation monitoring system, illustrating noisy data as described in FIG. 11 ;

FIG. 14 is a schematic plot of exemplary integrated microphone data and pressure sensor data from a blood circulation monitoring system, showing various features of the pulse wave shape, according to an exemplary embodiment of the invention; FIG. 15 is a schematic plot of a typical pulse wave shape and its second and fourth derivatives, illustrating a method of defining the beginning and end of systole and diastole intervals, according to an exemplary embodiment of the invention;

FIG. 16 is a flowchart of finding pulse wave velocity, according to an exemplary embodiment of the invention; FIG. 17 is a schematic plot of exemplary microphone data from two sensors with relative displacement, illustrating the method of FIG. 16; FIG. 18 is a flowchart for finding various blood circulation parameters from sensor data from a blood circulation monitoring system, according to an exemplary embodiment of the invention; and

FIG. 19 is another schematic plot of distributions of probability of the systolic point occurring at different pulse beats, using the method of FIG. 6.

DESCRIPTION OF EMBODIMENTS OF THE INVENTION

There is a need for an accurate, non-invasive, automated way to monitor a patient's circulatory parameters, such as blood pressure, systemic vascular resistance, and arterial compliance. In particular, there is a need to monitor blood pressure continuously, from beat to beat, accurately, non-invasively, and with minimum discomfort to the patient.

An aspect of some embodiments of the invention concerns using a microphone in a blood circulation monitoring system, to obtain pulse wave shape, by integrating the microphone signal over time. Optionally, a pressure sensor such as a load cell is also used to measure pulse wave shape, and the pressure sensor signal is combined with the integrated microphone signal to produce a pulse wave shape that is more accurate than either signal by itself. Because the microphone signal is generally proportional to the time derivative of the pressure, the microphone signal is more sensitive to fast changes in pressure. The pressure sensor may provide an absolute measure of pressure which does not drift as much in time as the integrated microphone signal. In addition, the pressure sensor signal may be less subject to motion artifacts than the integrated microphone signal.

An aspect of some embodiments of the invention concerns measuring pulse wave velocity with a blood circulation monitoring system, by measuring pulse wave shape at two points separated by less than 10 cm, or less than 5 cm, or less than 3 cm, optionally along a same artery. The difference in timing of the pulse wave at the two points is typically only a few milliseconds, comparable to the sampling time for the sensors measuring the pulse wave shape. Nevertheless, an accurate measure of the time difference between the two points may be obtained by using a comparator pin on a microcontroller for each of the sensors, which provides the time that the signal from the sensor crosses a particular value, for example zero, with a precision equal to the clock time of the microprocessor. Optionally the clock time is less than 100 microseconds, or less than 10 microseconds, or less than 1 microsecond, or less than 100 nanoseconds.

An object of some embodiments of the invention concerns a method of determining the systolic blood pressure of a patient, using a blood circulation monitoring system in which pressure is applied to an artery, but less than the systolic pressure, so that blood flow is not cut off completely. The pressure is gradually increased or decreased over several pulse cycles, while using a pressure sensor and/or a microphone to measure the amplitude and/or duration of the pulse beats in the artery at a location distal to the point where the pressure is applied. By measuring the pulse beats as a function of pressure, and extrapolating to higher pressures, the systolic pressure, at which the pulse would be cut off completely, can be estimated. The method has the potential advantage of being less uncomfortable for the patient than the usual oscillometric and auscultatory methods, in which the maximum applied pressure is greater than the systolic pressure, and blood flow is cut off completely.

An aspect of some embodiments of the invention concerns using pulse wave shape to measure rate and/or amplitude of breathing. Interthoracic pressure decreases during breathing, and arterial pressure may also decrease, due to a direct transmission of interthoracic pressure to the vascular tree, and/or to a decrease in stroke volume, caused for example by mechanical effects, reflex responses, and altered blood-gas tensions. Optionally, variations in blood pressure at the pulse frequency are filtered out, and/or the envelope of the variations at the pulse frequency is tracked in order to see the effect of breathing more clearly. In some embodiments of the invention, a breathing signal derived from the pulse wave is monitored in order to detect sleep apnea and other abnormal conditions, such as loaded breathing, airway obstructions, and snoring. The effect of breathing on the pulse wave signal may be greater for patients exhibiting these conditions.

An aspect of some embodiments of the invention concerns combining two or more data signals from a blood circulation monitoring system to determine one or more circulatory parameters, using a probabilistic or fuzzy logic analysis to determine a most likely value for each of the parameters, taking into account the uncertainty in the information provided by each data signal. For example, pressure sensor data and microphone data, taken while applying gradually decreasing pressure to an artery, are used to determine most likely values for systolic and diastolic pressure, taking into account the likelihood, for each of the signals, that a given pulse beat indicates the applied pressure is crossing the systolic point, or the diastolic point. Another example is the method used to combine the pressure sensor signal and the integrated microphone signal to find the likely pulse wave shape, as described above. When the two signals are generally in good agreement, the integrated microphone signal may be given more weight, at least for higher frequency fourier components of the pulse wave shape, because it is more sensitive to fast changes in pressure. But when the microphone signal shows momentary irregular bursts of high frequency components, they may be attributed to motion artifacts, and for those periods the pressure sensor signal may be given more weight.

An aspect of some embodiments of the invention concerns a method of segmenting data used to find blood circulatory parameters, for example pulse wave shape data, and data used in scaling events to find systolic and diastolic pressure. Optionally, segments of data signals which have characteristics that are very different from immediately earlier and later segments, by an appropriate measure, are labeled as noise segments, and disregarded or given lower weight for calculating blood circulation parameters. Optionally, other segments are grouped into homogeneous segments, and a representative pulse wave shape, for example an average pulse wave shape, is found for each homogeneous segment. Optionally, scaling events have their own segments, which will be referred to here as scaling segments, and the systolic and diastolic pressure is found for each scaling segment. Other parameters are optionally also found for each homogeneous segment, and for each scaling segment. Finally, the results found for each segment (for example, pulse shape for homogeneous segments, systolic and diastolic pressure for scaling segments) are displayed, together with the duration and time of day for each segment. Such a display of segmented data has the potential advantage that a physician can see at a glance the relevant results of many hours of data taking, for example for a hospital patient whose blood circulation parameters are being monitored. An aspect of some embodiments of the invention concerns a device for monitoring blood circulation parameters, such as blood pressure, worn around a part of the body such as a wrist, an upper arm, an ankle, or a thigh, in which blood flow in an artery is partially or completely occluded by applying pressure locally to an artery, for example a radial artery in the wrist. Such a procedure may be more comfortable for a patient than applying pressure all the way around the part of the body where the device is worn, as is done with a pressure cuff in prior art devices for monitoring blood pressure. The greater comfort in turn makes frequent repeated measurements of blood pressure more tolerable for the patient, allowing more timely and accurate monitoring. It should be noted that some embodiments of the invention have one or more of the following characteristics, which make may make it practical to use the invention to measure or monitor blood circulation parameters in a variety of settings where it may not have been possible to monitor blood circulation parameters previously, and which may save money and/or improve accuracy of results in traditional settings such as hospitals: 1) non-invasive measurement, does not require presence of trained medical personnel during use, for safety reasons;

2) automatic operation, does not require the presence of trained medical personnel to obtain accurate results, for example to determine systolic and diastolic pressure, and not subject to inconsistent results due to human error; 3) little or no discomfort to the patient, so measurements can be made frequently or continuously, and device can be worn by a patient continuously;

4) inexpensive, easily portable device that can be dedicated to continuous use by a single patient, and can be used at home, and while participating in sports or other daily activities, as well as in a doctor's office or in a clinical or hospital setting. Possible uses for some embodiments of the invention include: a. Individualizing choice and assessment of drug treatment for lowering cardiovascular risk. Different types of drugs, such as diuretics, beta-blockers, angiotensin converting enzyme, and calcium channel blockers have different effects on circulatory parameters. In assessing the effects of drugs on patients, parameters such as central systolic blood pressure, which may be derived from an accurately measured pulse wave shape, may be more important than more directly measured parameters such as peripheral systolic blood pressure, a point made in the CAFE Study cited above. b. Routine use of "pressure cardio gram" in addition to electrocardiogram (ECG),

Doppler blood flow, impedance cardiograms, pulse oximetry, and other standard measures, for diagnosis and monitoring. c. Replacement of invasive A-line monitoring in operating rooms and intensive care units, for monitoring of beat-to-beat blood pressure and pulse wave shape. A-line measurement is not used by many patients who could benefit from it, because it is risky, and it may be inaccurate because it can change the parameters it is measuring. d. Mobile ambulatory blood pressure monitoring, such as Holter monitoring. The use of data segmenting may be useful in removing noisy data segments, with motion artifacts, and allowing remaining data segments to be used for accurate determination of systolic and diastolic blood pressure, and other parameters of clinical interest that can be derived, for example, from pulse wave shape. e. Monitoring of vital signs at critical stages such as post-operative, emergency room, and other hospital departments. Such monitoring may be done as a supplement to an existing system used in an operating room and/or intensive care unit, or may be done with a low cost stand-alone system suitable for use by all patients in a hospital. f. Screening and routine checking for cardiovascular diseases. One-time measurement of blood pressure may be inaccurate, and even if accurate may reflect a combination of different cardiovascular diseases which have opposite effects on blood pressure, for example narrow arteries and low cardiac output, which can be revealed in accurate measurements of pulse wave shape. g. Clinical experiments by pharmaceutical companies, to evaluate new drugs and to follow up existing drugs. Beat-to-beat data from a portable continuously worn device can make clinical experiments shorter and more efficient, and can provide data on drug kinetics and adverse side effects. h. Telemedicine and remote monitoring of patients at home. Such monitoring, using wireless communication or regular checking by homecare professionals, may allow patients to remain at home who would otherwise have to be hospitalized. Such patients, and other patients at risk from cardiovascular disease, can benefit from increased accuracy, robustness to noise, and mobility and unobtrusiveness of monitoring devices. i. Monitoring patients at risk in daily life, for cardiovascular events. Existing systems monitor patients who are at cardiovascular risk. Such systems, for example, may allow a patient who feels like he may be suffering from a heart attack to call a monitoring center and transmit his ECG to the center. Transmitting a "pressure cardio gram" in addition to or instead of an ECG may provide additional information for diagnosis. j. Monitoring and diagnosis of patients during stress tests and during sport activities and exercise. Providing blood pressure and pulse wave shape data during stress tests or exercise may help patients not to exceed safe limits of activity. k. Detecting sleep cardiovascular disorders related to snoring and sleep apnea.

Since a patient's sleep may be disturbed by inflating a blood pressure cuff, making it impossible to measure cardiovascular disorders that occur only during sleep, it may be advantageous to be able to accurately monitor circulatory parameters by looking at pulse wave shape without applying external pressure to the artery being measured. Such monitoring, without disturbing sleep patterns, may be useful for diagnosing insomnia in a sleep lab, and for development of drugs to treat sleep disorders.

1. Assessing endothelial function. This can be done, for example, by prolonged occlusion of an artery, for example in the finger, and watching the subsequent recovery of blood flow. Assessing endothelial function may be important for determining the progress of vascular disease. m. Monitoring pilots, military or civilian, during flights, to make sure that g- forces and other stresses are not adversely affecting pilot performance during military maneuvers, for example, or to try to ensure that civilian or military pilots do not have an incipient undetected condition that could make it dangerous for them to fly.

Before explaining at least one embodiment of the invention in detail, it is to be understood that the invention is not necessarily limited in its application to the details of construction and the arrangement of the components and/or methods set forth in the following description and/or illustrated in the drawings. The invention is capable of other embodiments or of being practiced or carried out in various ways.

Referring now to the drawings, FIG. 1 illustrates a blood circulation monitoring system 100, which is held around the wrist, for example, of a patient, by a band 124. A wrist assembly 114 includes a pressure sensor 112, such as a load cell, and a vibration sensor 116 for oscillometric measurements, which are pressed against the patient's radial artery 128, with a controllable applied pressure that can be measured by sensor 112. The applied pressure, which is produced by an actuator shown below in Figs. 2 and 3, squeezes radial artery 128 against the distal part of the patient's radius bone 126, partially or completely occluding the flow of blood in the radial artery. In another embodiment of the invention, shown below in FIG. 4, the applied pressure is produced by a bag pumped up with air or another fluid. A force sensor 117, such as a load cell, and a contact microphone sensor 118, are pressed against radial artery 128 at a location distal to the location where sensors 112 and 116 are pressed against the radial artery.

It should be understood that, although sensor 112 is referred to as a pressure sensors, and sensor 117 is referred to as a force sensor, either sensor may actually be a force sensor, such as a load cell which directly measures total force independent of the contact area with the wrist, or a pressure sensor which directly measures pressure. If the contact area is constant, or depends only on the force, then the pressure may be a function of the force. As used herein, including in the claims, the term "pressure sensor" includes force sensors which directly measure force, since they may be used to infer the pressure, at least approximately.

Optionally, sensors 117 and 118 are pressed against radial artery 128 with a constant force, for example a force of 100 grams, or 50 grams, or 70 grams, or 150 grams, or 200 grams, independent of the applied pressure of sensors 112 and 116. The inventors have found that 100 grams, for example, provides enough force for good contact for making measurements with a load cell or a microphone, without impeding the flow of blood, or producing significant discomfort for the patient. It should be noted that 100 grams is much less than the force needed to cut off blood flow partially or completely. For example, if the force is applied over an effective area of 1 square centimeter, then a pressure of 100 mm of Hg, typical of the magnitude of diastolic or systolic blood pressure, corresponds to a force of 1360 grams.

Microphone 118 is used to detect Korotkoff sounds when making auscultatory measurements, and microphone 118 may also be used for measuring pulse wave shape and pulse wave velocity, as will be described below. A processing unit 110, optionally mounted on wrist assembly 114, controls the applied pressure, and optionally digitizes and analyzes data from the sensors. Alternatively processing unit 110 passes the sensor data to an external computer for data analysis. As will be described in detail below, processing unit 110 monitors clinically useful blood circulation parameters continuously, i.e. beat to beat, accurately and non- invasively, with little or no discomfort to the patient. Processing unit 110 also includes a power source both for its own use and for the wrist assembly, and optionally a display monitor and/or an interface to a computer.

In some embodiments of the invention, system 100 is held around a different part of the patient's body, and presses against a different artery, for example around the ankle pressing against the posterior tibial artery, around the upper arm against the brachial artery, around the thigh against the femoral artery. A similar system may also be used around the neck, for the carotid artery. If used for the carotid artery, the system preferably never presses with enough force to cut off blood flow completely, and optionally never presses with enough force to cut off blood flow even partially, for safety reasons, but is used, for example, to measure pulse wave shape and/or pulse wave velocity.

Pressure sensor 112, force sensor 117, vibration sensor 116, which may be a contact microphone, and microphone 118, may be, for example, similar to ones which can be obtained from Measurement Specialist, Inc., 1000 Lucas Way, Hampton, VA 23666, USA. For example, their load cell FS20, which goes up to 750 grams of force, is suitable to use for sensors 112 and 117. Their contact microphone with part number 1007079-1, for example, is suitable to use for vibration sensor 116 or microphone 118.

In some embodiments of the invention, as will be shown below in FIG. 4 for example, separate sensors 112 and 116 are replaced by a single sensor, which measures both pressure and vibrations. In the case shown in FIG. 4, the applied pressure is produced by an air bag, and the pressure and vibration sensor is located in processing unit 110, exposed to the pressure inside the air bag. Optionally, a piezo-resistive force sensor is used, such as the MPXV5050 sensor that can be obtained from Freescale Semiconductor, 7700 W. Parmer Lane, Austin, TX 78729-8084, USA.

FIG. 2 is a perspective overlay view of wrist assembly 114 and band 124, surrounding the patient's wrist, with sensors 116, 117, and 118 adjacent to radial artery 128. Band 124 is optionally made of a soft material such as leather or silicone, with a skeleton of metal wires, optionally a shape memory alloy such as NiTi which will cause band 124 to maintain its shape once it is fitted. Band 124 is optionally locked in place by a locking mechanism 282, not visible in FIG. 2 because it is on a side of wrist assembly 114 that is hidden. Locking mechanism 282 optionally locks and unlocks by moving in a direction shown by arrow 284.

Optionally, an actuator 291 applies a controllable force to sensor 116, pressing it against radial artery 128 when using the oscillometric and auscultatory methods to measure systolic and diastolic pressure, as will be described below. Actuator 291 is optionally controlled, for example, by a motor 286, inside wrist assembly 114. Alternatively, motor 286, or another kind of actuator inside wrist assembly 114, controls the pressure applied at sensor 116 by tightening or loosening band 124, for example by pulling the free end of band 124 into wrist assembly 114.

Optionally, wrist assembly 114 has a mechanical connector 294, such as snaps, for attaching processing unit 110, and an electrical connector 296, to obtain electric power from processing unit 110. Optionally, processing unit 110 is a processing unit used in a prior art system for blood pressure monitoring using an inflatable mini-cuff that goes around the wrist, and snaps 294 and electrical connector 296 are designed to fit such a commercially available processing unit. Optionally, software used in processing unit 110 is modified for use with wrist assembly 114, but the hardware of processing unit 110 is identical to a commercially available processing unit for such a mini-cuff blood pressure monitoring system.

FIG. 3 is a cross-sectional view of a blood circulation monitoring system 300, similar to system 100, including a wrist assembly 320, and a processing unit 330. Processing unit 330 includes an actuator 342, for example a stepper motor, which presses pressure sensor 350 (similar to pressure sensor 112 in FIG. 1), and vibration sensor 352 (similar to vibration sensor 116 in Figs. 1 and 2) against radial artery 128, squeezing the artery against radius bone 126. Optionally, sensors 350 and 352 are contained in a dome-shaped structure that allows the sensors to change orientation, for example with two degrees of freedom. When the sensors are pressed against the skin, they tend to become oriented perpendicular to the skin surface above the radial artery. As a safety feature, there is optionally a spring 340 above actuator 342, which starts to contract if actuator 342 presses with too much force, limiting the force that actuator 342 can exert on artery 128, in case the software controlling the applied pressure fails to limit the force. Additionally or alternatively, the force is limited by a limit in the length by which actuator 342 can expand.

Optionally, there is a display unit 332 on the upper surface of processing unit 330, which displays one or more parameters of blood circulation obtained by analyzing data from the sensors, as will be described below. The display may be alphanumeric, graphic, or both. Optionally, processing unit 330 also includes a control panel 336, with, for example, buttons for controlling the operation of system 300, including an on/off button or switch.

Processing unit 330 also includes electronic circuitry 360, which includes a microprocessor for controlling the applied pressure and analyzing data from the sensors, and interfaces to the sensors and actuator, as will be described in more detail in FIG. 5. FIG. 4 is a cross-sectional view of a blood circulation monitoring system 400, with a wrist assembly 420 and a processing unit 440. System 400 is similar to system 300, but with a fluid-filled bag, for example an air-bag or cuff 442, exerting pressure on radial artery 128 when using the oscillometric and auscultatory methods, instead of a stepper motor or actuator. Air-bag 442 is pressurized by a pump 440, and is deflated, reducing the pressure on the radial artery, by an electrically controlled relief valve 444. Optionally, air-bag 442 is in the form of a cuff, going completely around the wrist, and inflating the cuff causes it to press against the radial artery. Alternatively, air-bag 442 does not go completely around the wrist, but is located above the radial artery, and held against the radial artery by a strap, for example. Pump 440, and optionally relief valve 444, are under the control of a microprocessor in an electronics board 460, similar to electronics board 360 in system 300, which can therefore control the pressure applied to the radial artery as a function of time. For example, when using the oscillometric and auscultatory methods, pump 440 may pump up air-bag 442 to a high enough pressure to occlude artery 128 completely, i.e. higher than the systolic pressure, while keeping valve 444 closed. Once the maximum pressure is reached, pump 440 is turned off, and valve 444 is opened enough so that the pressure of air-bag 442 starts slowly decreasing, over many pulse cycles. Alternatively, valve 444 is always open enough to produce that rate of decrease in pressure, but pump 440 works hard enough to increase the pressure of air-bag 442, in spite of the losses through the valve. When the maximum pressure is reached, pump 440 turns off, and the pressure starts to decrease. In this case, valve 444 need not be controllable by electronics board 460.

A mechanical limiter 448 optionally limits how far down air-bag 442 can extend from wrist assembly 420, and hence how much force air-bag 442 can apply to artery 128.

Sensor 450 is exposed to the pressure of air-bag 442, and measures the applied pressure exerted by air-bag 442 on the artery, as well as the vibrations caused by the pulse in the artery when the applied pressure is less than the systolic pressure.

Like processing unit 330 in system 300, processing unit 430 optionally has a display 432, and a control panel 436.

FIG. 5 shows a block diagram 500 of the hardware of system 400. Dashed rectangles distinguish the elements that are part of a pressure module 502 and a distal sensor module 504, both of which are in wrist assembly 420, as well as the elements that are part of electronics board 460.

Pressure module 502 includes air-bag or cuff 442, pump 440, and release valve 444. Pump 440 and, optionally, release valve 444, are connected, through ports 520, to a microprocessor 522, which controls the pressure in air-bag 442. A digital to analog converter 524 transforms the digital control signals produced by microprocessor 522 into voltages that control the rate of pumping of the pump and the degree of opening of the valve.

Sensor module 504 includes force sensor 117 and microphone sensor 118. Microphone 118 is optionally connected to a low pass filter 526, to filter out high frequency noise in the signal, for example above 10 Hz or above 20 Hz or above 35 Hz.

Additionally or alternatively, microphone is also connected to a high pass filter 527, to filter out a low frequency portion of the signal, for example below 10 Hz or 20 Hz or 35

Hz. Filters and 526 and 527 are optionally connected to two amplifiers 528, one for the low frequency portion of the microphone signal and one for the high frequency portion of the microphone signal. Amplifiers 528 are optionally connected to microprocessor

522 through an AC to DC converter 530. Pressure sensor 450 is optionally a digital sensor, with its own amplifier and AC to DC converter, and is directly connected to port

520 of microprocessor 522. Optionally, data from pressure sensor 450 is divided into low frequency and high frequency components by software running on microprocessor 522. Alternatively, pressure sensor 450 is an analog sensor, and uses a low pass filter and/or a high pass filter each with its own amplifier, connected to AC to DC converter 530, like microphone 118. Alternatively, microphone 118 is digital and has its own amplifier and AC to DC converter, and is connected to microprocessor 522 through ports 520, and optionally uses software to separate low frequency and high frequency components.

Force sensor 117 is optionally connected to a low pass filter 531 and a high pass filter 532, which are connected to two amplifiers 534, one for a low frequency component and one for a high frequency component of the data signal from sensor 117. Sensor 117 optionally uses two filters and amplifiers because its output includes a DC portion that is sensitive to the slow change in pressure during a scaling event, and a higher frequency portion that measures the pulse wave shape without following the long term trend. Alternatively, sensor 117 has a single output, with a single amplifier and filter, and higher and lower frequency portions are separated by software. The amplifiers are connected, via analog to digital converters 530, to microprocessor 522. Optionally, sensor 117 is a digital sensor with its own amplifier and AC to DC converter, and connects to microprocessor 522 through ports 520.

Optionally, a sensor input interface 506 provides an interface for connecting other sensors to microprocessor 522, for example one or more of an electrocardiogram (ECG), an impedance cardiogram (ICG), an impedance plethysmography (IPG) sensor, an oximeter based for example on photoplethysmography (PPG), and an ultrasound microphone. Any of these sensors may be digital or analog, with an appropriate connection to microprocessor 522. Optionally, instead of or in addition to interface 506, one or more of these other sensors is included in system 400.

Optionally, microprocessor 522 includes a signal evaluation board 536, which is programmed to analyze the sensor data to find various blood circulation parameters, as will be described below. Alternatively, microprocessor 522 does not use special hardware to analyze the data, but has software loaded into its memory which is used to analyze the data, or microprocessor sends the data to an external computer, for example an ordinary personal computer, which analyzes the data, and optionally returns results for microprocessor 522 to display on display 432. Whether or not an external computer does the data analysis, electronics board 460 optionally includes an interface 538 to an external computer, for example an RS 232 connector or a USB connector. Alternatively or additionally, electronics board 460 includes a wireless communication link 539.

There are also optionally one or more batteries 540 to serve as a power source, for example one or two 1.5 volt batteries, and/or a 9 volt battery, optionally rechargable. It may be advantageous to have separate batteries for the pump, which may consume more power but only runs intermittently, and the microprocessor, sensors and amplifiers, which may consume less power but run continuously, so that the battery for the pump can be replaced, or recharged, without turning off the microprocessor. Alternatively or additionally, there is a connection to an external power supply, not shown in FIG. 5, for directly powering the system instead of using batteries, or for recharging the batteries without removing them. Optionally there is a voltage regulator 542, to avoid fluctuations in voltage, and/or to provide power to electronics board 460 and to the sensors at a different voltage than provided by the batteries. Voltage regulator 542 may also compensate for any changes in voltage of the batteries as they are used up.

Optionally there is an on-off switch or button 544, included for example in control panel 436, and a power-on indicator 546, such as an LED.

It should be noted that a block diagram of the hardware of system 300, or system 100, might differ from block diagram 500 only in that pressure module 502 would comprise an actuator controlled by microprocessor 522, rather than a pump and a relief valve, and the software used by microprocessor 522 for this purpose might be different. For example, it might be possible for the microprocessor to directly command the actuator to extend to a certain position, while in the case of the pump, the microprocessor might only be able to turn it on and wait for the pressure to build up to a desired value, or the microprocessor might directly command only the rate of pumping, and hence the rate of increase in pressure. Also, in the case of systems 100 and 300, there might be a separate pressure sensor and vibration sensor in the pressure module, each with its own amplifier and connection to the microprocessor in the electronics board, rather than a single pressure sensor 450 used for both. FIG. 6 is a flowchart 600 for a "scaling event," which is a procedure for measuring systolic and diastolic pressure in which pressure is applied to the artery and gradually deceased, and sensor data is used to determine when the applied pressure goes past the systolic and diastolic points. The sensor data can include, for example, vibration data (in the oscillometric method), microphone data for detecting Korotkoff sounds (in the auscultatory method), pulse oximetery data, and other sensor data. For the procedure outlined in FIG. 6, more than one kind of sensor data is used, and the results are combined to obtain a measure of systolic and diastolic pressure that may be more accurate than could be obtained by using data from any one sensor. At 602, a scaling event in initiated. This is done, for example, when the estimated uncertainty in systolic and diastolic pressure, found by monitoring the pulse wave shape since the last scaling event, exceeds a chosen threshold. Optionally, the threshold in uncertainty depends on the subjective discomfort to the patient of a scaling event, which can be uncomfortable because it temporarily stops blood flow in the artery completely, and a higher threshold in uncertainty is used for patients who find scaling events more uncomfortable. Optionally, a lower threshold in uncertainty in systolic and diastolic pressure is set for "mini- scaling" events, described in Figs. 9 and 10, than for scaling events. Mini- scaling events cause less discomfort for the patient than scaling events, because they do not cut off blood flow in the artery completely, but use extrapolation to estimate the systolic pressure. On the other hand, mini-scaling events may give less accurate results for systolic pressure, and eventually, even if one or more mini-scaling events are performed, the uncertainty in systolic pressure may grow to a level that a scaling event is initiated.

At 604, the applied pressure on the artery is increased, for example using an actuator or an air-bag, as described in Figs. 1-4. Optionally, data from one or more of the sensors is monitored as the pressure is increased, to determine whether the applied pressure exceeds the systolic pressure, at 606. Sensors 117 and 118, distal to the point where pressure is applied, may be particularly useful in this regard, since when the systolic pressure is exceeded, the blood flow in the artery is cut off completely, and the signals from these distal sensors go nearly to zero. Once the systolic pressure is exceeded, the applied pressure stops in increasing, and starts to decrease gradually, at 608. It is not necessary to stop the applied pressure from increasing precisely at the point where the blood flow is cut off completely, but, in this embodiment, the applied pressure stops increasing when it is reasonably certain that that point has been passed. Optionally, the applied pressure is allowed to continue to increase further by some amount after the point where the blood flow is cut off completely, for example by 10 mm Hg, or 20 mm Hg, or 50 mm Hg, or by 5% or 10% or 20% or 40%. Allowing the applied pressure to increase further has the potential advantage that the entire transition through the systolic pressure may be seen when the applied pressure is gradually decreased, allowing a more accurate determination of the systolic pressure.

In some embodiments of the invention, the sensors are not monitored while the applied pressure is increased, and the applied pressure is increased up to a previously chosen level, believed to be definitely greater than the systolic pressure, for any patient, or at least for that patient. A potential advantage of monitoring the sensors while the applied pressure increases, and stopping the increase just when the systolic pressure is exceeded, or a little after the systolic pressure is exceeded, is that the maximum applied pressure will not be much greater than necessary, and the blood flow will not be cut off for a much longer time than necessary, reducing discomfort to the patient.

The rate of decrease is chosen so that it will take many pulse cycles, for example about 60 pulse cycles, or between 40 and 80 pulse cycles, to decrease to zero. Decreasing the applied pressure too slowly may cause more discomfort to the patient, since the blood flow is cut off for a longer time, and may even be dangerous if the blood flow is cut off for too long a time. Decreasing the applied pressure too quickly will result in a less precise measurement of systolic and diastolic pressure. Decreasing the applied pressure too slowly may also result in a loss of precision, if the measurement takes such a long time that the systolic or diastolic pressure may change significantly during that time.

As the applied pressure decreases, sensor data is recorded at 610. In some embodiments of the invention, the recorded data is only analyzed after the applied pressure has decreased to zero, completing the scaling event. In other embodiments, including the embodiment shown in FIG. 6, the systolic and diastolic pressures are found by analyzing the data in real time, which has the potential advantage that the results are available sooner, and that the applied pressure can be quickly released once the diastolic point is definitely passed. In either case, the same steps are used in analyzing the data. At 612, the sensor data is examined, and at 614, a decision is made as to whether the applied pressure has definitely passed below the systolic pressure. This is determined, for example, by looking at the data from sensors 117 and/or 118, located distal to the point at which pressure is applied to the artery. If there is a substantial signal from these sensors, or there has been a substantial signal for at least a few pulse beats (so it is clear that the signal is not just noise), then the systolic point has been passed. Optionally, the data up to that point is analyzed at 616 and 618, to calculate the systolic pressure more precisely, in parallel to continuing to monitor the data for the diastolic point at 620 and 622. At 616, a probability is found that each pulse beat is the systolic point. An exemplary procedure for doing this will be described in Figs. 7 and 8. At 618, the most likely systolic pressure is found by taking the pulse beat that is most likely to be the systolic point, or taking the mean or median pulse beat in the probability distribution, and finding what the applied pressure was at the time of that pulse beat. The error bars in the systolic pressure are found by using points on the probability distribution for each pulse beat to be the systolic point, for example one sigma points or two sigma points, and seeing what the corresponding range in applied pressure is.

In some embodiments of the invention, once it is believed that the systolic point has been passed, the pressure is increased again above the systolic point, and then decreased again below the systolic point, one or more times, recording data as the systolic point is passed, optionally both while the pressure is increasing and while the pressure is decreasing. A scaling event that uses this procedure is referred to as a "super- scaling event." Optionally this procedure is only done, or it is only repeated, if it is believed that the one or more measurements already made of the systolic pressure are not as accurate as they could be, for example because the data is unusually noisy (e.g. above a noise threshold), or seems inconsistent, or implies a systolic pressure that is very different from an expected value (e.g. based on history or on one or more thresholds). In some embodiments of the invention, the pressure is decreased and increased repeatedly over a range that is believed in advance to include the systolic pressure, but it is not necessarily verified in real time, by analyzing the data, that the systolic pressure has been passed. A super- scaling event has the potential advantage over an ordinary scaling event in that a more accurate value of systolic pressure may be obtained, but an ordinary scaling event has the potential advantage that it may cause less discomfort to the patient.

Super- scaling events may be particularly useful during surgery, when accurate measurement of systolic pressure is particularly important, and when the patient is unconsciousness and will not feel any discomfort in any case. Similarly, super-scaling events may be useful for a patient in critical condition, when accurate measurements of systolic pressure may be especially important, and the patient may not be conscious. Optionally, in such situations, e.g. with an unconscious patient or a patient in danger, super- scaling events are triggered automatically whenever other diagnostics indicate a possibility that the systolic pressure has dropped significantly. Meanwhile, sensor data continues to be recorded at 620, and is optionally examined at 622 to decide, at 624, whether the diastolic point has been passed. Indications for passing the diastolic point are described below in FIG. 7. The goal at 624 is not necessarily to obtain a precise value for the diastolic pressure, but only to determine whether, with a high degree of confidence, the diastolic point has already been passed. If the diastolic point has not yet been passed, and (at 626) the applied pressure has not yet reached zero, then data continues to be recorded at 620, as the applied pressure decreases. If the diastolic point is believed to be definitely passed, then, at 628, the applied pressure is quickly reduced to zero. Once the applied pressure has reached zero, whether or not it was determined at 624 that the diastolic point was definitely passed, then the recorded data is examined more carefully, at 630 and 632, to find a precise value for the diastolic pressure.

At 632, a probability is calculated for each pulse beat being the diastolic point, as will be described below in Figs. 7 and 8. At 634, the most likely diastolic pressure is found by taking the pulse beat with the greatest probability of being the diastolic point, or taking the mean or median of pulse beats in the probability distribution for the diastolic point, and seeing what applied pressure corresponds to that point. The error bars in the diastolic pressure are found by using points on the probability distribution for each pulse beat to be the diastolic point, for example one sigma points or two sigma points, and seeing what the corresponding range in applied pressure is. Optionally, a correction is made to the systolic and/or diastolic pressure, due to the finite stiffness of the artery wall, which may vary depending on the degree of calcification of the wall, for example. Because of the stiffness of the artery wall, the applied pressure needed to occlude the artery, at any time during the pulse cycle, is greater than the blood pressure inside the artery at that time. The additional applied pressure needed to overcome the stiffness of the artery wall is optionally estimated, using any known method of measuring artery stiffness, and subtracted from the applied pressure at the systolic and/or diastolic time, to obtain a corrected systolic and/or diastolic pressure. The stiffness of the artery wall is estimated, for example, using the local pulse wave velocity, measured as described below in the description of FIG. 17. Although the pulse wave velocity may depend on a different component of the stiffness tensor, or a different combination of components, than the correction to the systolic and diastolic pressure, it may be possible to find a relationship between them using a model for arterial wall stiffness.

In some embodiments of the invention, a scaling event is used to measure one or both of the systolic and diastolic pressure, and a correction is made due to the stiffness of the artery wall as described above, even without performing some of the other actions described in flow diagram 600, for example even without combining probability distributions from different data signals.

FIG. 7 shows a plot 700 of the applied pressure 702 as function of time 704, as measured for example by pressure sensor 112, as well as plots of six other data signals, all with the same time axis 704: 1) a plot 706 of the signal from the vibration sensor 116 at the same location as the applied pressure, including an envelope 708 showing the maximum of each pulse beat; 2) a plot 710 of the detrended signal from pressure sensor 112, or pressure sensor 350 in FIG. 3, showing only the variation in pressure associated with the pulse beats; 3) a plot 712 of the signal from microphone 118 located distal to the applied pressure, but filtered to provide only the lower frequency components, for example below about 35 Hz, and including an envelope 714 showing the maximum of each pulse beat; 4) a plot 716 of the higher frequency components, for example above about 35 Hz, of the signal from microphone 118; 5) a plot 718 showing a DC portion, with the pulse beats greatly attenuated, of a signal from force sensor 117, distal to the applied pressure; and 6) a plot 720 of an AC portion of the signal from force sensor 117, predominantly showing the pulse beats.

Considering only plot 706, the most likely pulse beat for the systolic point is a pulse beat 722, which has a height equal to 60% of the height of a highest pulse beat 724, and occurs at time 728 before pulse beat 724. The most likely pulse beat for the diastolic point is a pulse beat 718, which has a height equal to 80% of the height of pulse beat 726, and occurs at time 730, after pulse beat 724. Note that as the applied pressure decreases from its initial high value, the amplitude of the pulse beats detected by vibration sensor 116 initially increases, because more blood can flow through the artery as the applied decreases. However, as the applied pressure starts to approach the diastolic pressure, then decreasing the applied pressure further does not increase the blood flow very much. At the same time, decreasing the applied pressure decreases the coupling between the vibration sensor and the artery, so the amplitude of the signal starts to go down with decreasing applied pressure.

The peaks of the pulse beats in plot 710 show a similar trend to curve 708 of the peaks of the pulse beats in plot 706, and are optionally used in addition to, or instead of, curve 708 and plot 706. Optionally, the most likely pulse beat for the systolic point is, for example, the fourth pulse beat for which the pulse beat signal, as measured by sensor 112 (or sensor 350), exceeds a certain threshold.

Envelope 714 in plot 712 is expected to remain close to zero when the applied pressure is greater than the systolic pressure, and only starts to rise at a time 736, which is the most likely time for the systolic pressure, based on the data in plot 712. Envelope 714 reaches a maximum at a time 738, which is the most likely time for the diastolic point, based on plot 712. At lower applied pressure, curve 714 falls, because microphone 718 is not in as good contact with the skin.

The high frequency part of the microphone signal in plot 716, which shows the Korotkoff sounds, is expected to be substantially zero when the applied pressure is greater than the systolic pressure, but suddenly rises to its maximum value at a time 740, which is the most likely time for the systolic point, based on plot 716. The amplitude of the signal associated with each pulse beat has another, lower, local maximum at a time

742, which is the most likely time for the applied the diastolic point, and corresponds to the stage 4 and stage 5 Korotkoff sounds which are characteristic of the diastolic point.

In plot 718, the signal is expected to be nearly zero when the applied pressure is greater than the systolic pressure, and starts to rise starting at time 744, which is the most likely time for the systolic point, based on plot 718. An inflection point in the signal (ignoring the ripple due to individual pulse beats) at time 746 indicates the most likely time for the diastolic point.

In plot 720, the envelope of the signal is expected to be nearly zero for applied pressure greater than the systolic pressure, and starts to rise at time 748, which is the most likely time for the systolic point, based on plot 720. The envelope of the signal reaches a maximum shortly thereafter, and remains close to its maximum for the rest of the scaling event, because, in contrast to pressure sensor 112, force sensor 117 is pressed against the skin with a nearly constant pressure. For this reason, the signal in plot 720 is optionally not used to estimate the diastolic point. Knowing at which pulse beats the systolic and diastolic points occur, the systolic and diastolic pressure may be found by seeing what the applied pressure was at those times. For example, if the systolic point occurs at time 728, and the diastolic point occurs at time 730, then the systolic pressure 732 and the diastolic pressure 734 can be read off plot 700. FIG. 8 shows a plot 800 of the probability 802 that each pulse beat is the systolic point, based on the data from microphone 118 plotted in plot 712 of FIG. 7, and a plot 804 of the probability 806 that each pulse beat is the systolic point, based on the data from vibration sensor 116 plotted in plot 706 of FIG. 7. The time axes 704 for plots 800 and 804 correspond to the time axes 704 in FIG. 7. Note that probability 802 goes rapidly to zero for pulse beats much earlier than pulse beat 736, has a peak at time 736, and goes more gradually to zero for pulse beats later than time 736. The systolic point could occur later than time 736, for example, if the signal at time 736 were due to noise, or due to a pulse from another nearby artery that does not have pressure applied to it. Probability 806 has a peak value at time 722, and falls off gradually for pulse beats before and after that, based on the relative height of those pulse beats in plot 706 to the height of the pulse beat at time 724, the highest pulse beat.

Plot 808 schematically shows the overall probability 810 that a given pulse beat is the systolic point, taking into account data from both vibration sensor 116, plotted in plot 706, and microphone 118, plotted in plot 712. Probability 810 is optionally calculated by combining probabilities 802 and 806 in a usual way of combining Bayesian or conditional probabilities, for example by multiplying the two probabilities together and normalizing. Note that probability distribution 810 is narrower than either distribution 802 or 806, and hence would produce less spread in the value of the systolic pressure if projected onto the vertical axis of plot 700. In some embodiments of the invention, additional sensor data, for example the data in the other plots in FIG. 7, is used to find the overall probability distribution 810 of the systolic point, resulting in a still narrower distribution. Optionally, data from other types of sensors is also used, for example pulse oximetry sensors, PPG sensors or IPG sensors.

Similar methods are used for determining the probability distribution of a given pulse beat being the diastolic point, based on the data for each sensor, and for combining these probability distributions in a Bayesian sense into an overall probability that each pulse beat is the diastolic point.

Further details on how the probability distributions are found, for an exemplary embodiment of the invention, are given below in the Appendix.

FIG. 9 is a flowchart 900 for a "mini- scaling" event. As mentioned above in the description of FIG. 6, a mini-scaling event is a procedure, like a scaling event, which is used to find the systolic and diastolic pressure, by applying a changing pressure to an artery, and looking at the resulting sensor data on the pulse beat in the artery. However, unlike a scaling event as described in FIG. 6, in a mini-scaling event the applied pressure always remains below the systolic pressure, so that the blood flow in the artery is never cut off completely. This results in less patient discomfort, at the possible cost of a less accurate measure of the systolic pressure.

Mini- scaling events may be particularly useful for regular routine monitoring of blood pressure when a patient is sleeping, since great accuracy may not be needed for routine monitoring if there is no particular reason to suspect a change in blood pressure, while the patient's sleep may be less likely to be interrupted if the procedure is less uncomfortable. This may be particularly important when monitoring a patient with a sleep disorder.

At 902, the mini-scaling event is initiated, for example because uncertainty in the values of systolic and diastolic pressure has grown greater than a chosen threshold, since the last scaling or mini-scaling event. The threshold is optionally greater for patients who find the procedure more uncomfortable, and, since a mini-scaling event is generally less uncomfortable than a scaling event, the threshold for initiating a mini-scaling event is optionally lower than the threshold for initiating a scaling event. Optionally, scaling events are not used at all, and mini- scaling events produce adequate precision in measuring the systolic and diastolic pressure.

At 904, an estimated minimum applied pressure is chosen, for example a pressure that is believed to be definitely below the diastolic pressure, or a pressure of zero, and a rate of increase of applied pressure is chosen. The rate of increase is, for example, similar to the rate of decrease of applied pressure used in scaling events, with the applied pressure increasing from the initial pressure to the estimated systolic pressure in about 40 or 50 pulse cycles, or between 30 and 60 pulse cycles, although, as noted above, the applied pressure never actually reaches the systolic pressure. At 906, the applied pressure is increased to the chosen minimum, and at 908, the applied pressure is further increased at the chosen rate. At 910, data is recorded from the sensors. Optionally, it is determined in real time, at 912, whether the diastolic point has been passed, using for example the criteria described for determining the diastolic point in FIG. 7, and the diastolic point is recorded in 914. Alternatively, the data is only analyzed to determine the diastolic point after the mini- scaling event, but knowing the diastolic point in real time has the potential advantage that it may make it possible to predict the systolic pressure more precisely in real time. In either case, after the applied pressure exceeds the diastolic pressure, data from the sensors is recorded at 916, and a prediction of the systolic pressure is made at 918. Optionally, this prediction is made by looking at the trend of data with increasing applied pressure in real time, particularly the trend of data that is sensitive to the value of the systolic pressure, such as the data from force sensor 117 and microphone 118 distal to the point where pressure is applied to the artery, or other sensors distal to the point where pressure is applied. Generally, the variation with pulse cycle of such data will go essentially to zero at the systolic point, and, for example, a linear projection of the vanishing point of pulse cycle variation with increasing applied pressure is used to estimate the systolic pressure, as described below in FIG. 10. If the prediction is made based on the trend of the data is real time, then it is optionally updated repeatedly as the applied pressure is increased. Additionally or alternatively, the prediction of systolic pressure is based on the previous measurement of systolic pressure using a scaling or mini-scaling event, and/or on estimates of systolic pressure based on data taken since the last scaling or mini-scaling event, for example based on pulse shape measurements, and/or on pulse wave velocity measurements, as will be described below. Additionally or alternatively, historical data on the systolic pressure of this patient and its variation, or historical data on systolic pressure in a larger population, are used to estimate systolic pressure. If no real time data is used, then optionally the prediction of systolic pressure is not updated during the mini-scaling event. Updating the prediction, based on real time data, has the potential advantage that the systolic pressure can be estimated more accurately as the applied pressure increases, allowing a better optimized maximum applied pressure to be used.

At 920, a decision is made as to whether the applied pressure is close enough to the systolic pressure, so that it should not be increased any further. The maximum applied pressure is, for example, a certain percentage of the latest predicted systolic pressure, for example, 60%, 70%, 80%, 90%, or higher, lower, or intermediate values. Choosing a higher percentage of the predicted systolic pressure for the maximum applied pressure will produce a more accurate estimate of systolic pressure, at the cost of possibly greater discomfort for the patient during the mini-scaling event, because the blood flow in the artery will be more nearly cut off. A higher maximum applied pressure, relative to the predicted systolic pressure, may be used, for example, if there is more uncertainty in the patient's systolic pressure based on previous data, or if the systolic pressure tends to be unusually unstable for this patient, or if it is particularly important to get an accurate measure of systolic pressure. A lower maximum applied pressure may be used, for example, if the patient is particularly sensitive to the discomfort of a high applied pressure, or if the systolic pressure is known fairly well already.

Once the maximum applied pressure has been reached, the final predicted systolic pressure is optionally recorded, at 922, as the estimated systolic pressure. This is done, for example, if the predicted systolic pressure had been updated in real time as the applied pressure increased, based on the sensor data. If the predicted systolic pressure had not been updated in real time during the mini-scaling event, then the recorded data is analyzed after the maximum applied pressure has been reached, to make a best estimate of the systolic pressure, based on trends in the sensor data with increasing applied pressure, as described below in FIG. 10, and the estimated systolic pressure is then recorded. The diastolic pressure is also estimated from the data, using criteria similar to those described for FIG. 7, and recorded, if this had not been done earlier in real time. The applied pressure is released at 924. If the systolic and/or diastolic pressure is only estimated after the data is taken, rather than in real time, then optionally the applied pressure is released before the systolic and/or diastolic pressure is estimated and recorded. Optionally, a correction is made to the systolic and/or diastolic pressure, due to the finite stiffness of the artery wall, as described above for FIG. 6.

Optionally, the procedure described in flowchart 900 is repeated one or more times, and the estimated systolic pressures (and optionally diastolic pressures) obtained each time are combined, for example by taking an average or a weighted average, to obtain a more accurate value for systolic pressure (and optionally the diastolic pressure). Optionally, the estimated values of systolic pressure obtained each time are used when deciding the maximum applied pressure to use when the procedure is repeated, which may improve the accuracy of the estimated systolic pressure when the procedure is repeated. Repeating the mini- scaling procedure may be particularly useful, since 1) the mini-scaling procedure may give a less accurate estimate of systolic pressure than a scaling event, each time it is done, but repeating it may provide improved accuracy; and 2) a mini-scaling event is less uncomfortable for the patient than a scaling event, so it is not so uncomfortable to repeat it one or more times.

In some embodiments of the invention, the applied pressure in mini-scaling events is initially set at a maximum value, believed to be below the systolic pressure, for example at 60%, 70%, 80% or 90% of the systolic pressure, and gradually decreased while data is taken. The data is then used to estimate the systolic pressure, based for example on a linear extrapolation, as described below in FIG. 10. However, starting at low applied pressure and increasing the applied pressure has the potential advantage that the maximum applied pressure can be optimized in real time, rather than chosen in advance, potentially allowing a better trade-off between precision of the estimate of the systolic pressure, and patient comfort.

FIG. 10 is a plot 1000 of a signal 992 from microphone sensor 118, as a function of time, when the applied pressure is being increased up to the systolic pressure. FIG. 10 illustrates how the systolic pressure may be predicted in real time during a mini-scaling event, or estimated afterward from the recorded data. A straight line 990 is shown fitted to the minima of the signal at each pulse beat, starting at pulse beat 996, after the applied pressure exceeds the diastolic pressure. The straight line reaches zero at point 994, which corresponds to the systolic pressure. It should be noted that all of the minima of the signal at each pulse beat are close to line 990. Even if the applied pressure were not increased all the way to the systolic pressure, but were increased only to an intermediate pressure between pulse beat 996 and point 994, fitting the points to a straight line would produce nearly the same line 990, which would reach zero at nearly the same point 994. So the systolic pressure can be estimated or predicted very well with this method, without having to increase the applied pressure all the way to the systolic pressure, and without cutting off the blood flow completely. It should be noted that the maxima of signal 992 at each pulse beat could be used in the same way as the minima, except that in the data plotted in FIG. 10, the maximum is saturated for all except the last few pulse beats.

Optionally, the maxima or minima of the data for each pulse beat are fitted to a curve with more free parameters than a straight line. Optionally, features of the data for each pulse beat other than the maxima or minima, for example some kind of integral or derivative of the data for each pulse beat, is fitted to a straight line or another curve.

FIG. 11 is a flowchart 1100 for a method of segmenting blood circulation data. Segmenting the data is advantageous when a patient is monitored for a long period of time, and it would be burdensome for a physician or other medical personnel to examine all the data. Instead, the data is broken up into homogeneous segments, and for each segment, the duration of the segment is recorded, together with average or typical data from that segment, for example a typical pulse wave shape. In addition, noise segments of the data, for example due to motion artifacts, can be separated from segments of good data. At 1102, the data is divided into data segments, which comprise data taken with no applied pressure, between scaling events, and scaling segments which comprise data from scaling events (including any mini- scaling events). This information is optionally recorded when the data is recorded, or, if not, is apparent from looking at the applied pressure as a function of time. At 1104, the next segment is examined, and at 1106, it is decided whether it is a data segment or a scaling segment. If it is a data segment, then the pulses are identified at 1108, and distinguished from noise between pulses that may have amplitude as great as the pulses for some sensor data. Pulses are distinguished from any high amplitude noise found between pulses, for example, by the fact that they have an expected shape, and are nearly periodic at a pulse frequency.

The first pulse is looked at, at 1110. At 1112, a determination is made as to whether the data for this pulse is atypical, for one or more sensors, and therefore likely to include significant noise, due to motion artifacts for example. In the case of the first pulse, this determination is made, for example, by comparing the pulse shape to a general expected pulse shape for this sensor, or to pulse shapes from previous segments from the same patient. For later pulses, the pulse may be compared to a predicted pulse shape based on earlier pulses in the same segment, as will be described. If it is determined that the data for this pulse is atypical, for at least one sensor, then the pulse is labeled as noise, at least for that sensor, at 1114. A running average of the pulse shape for the data from each sensor, for example an auto-regressive moving average, is then made at 1116, and a prediction is made for the next pulse, for the data from each sensor. Optionally, an optimal or suitable number of pulse beats to include in the regressive moving average is determined empirically, and possibly adjusted in real time. A pulse shape that has been labeled noise is optionally excluded from the running average and prediction, at least for the sensor data that has been labeled noise. For other sensor data, the pulse may be typical, and is optionally included in the running average and prediction for that sensor data. At 1118, the data is examined to see if any more pulses remain. If so, the next pulse is examined at 1120, and compared to the prediction made at 1116. Control then returns to 1112, where it is decided whether, for the data from each sensor, the new pulse is atypical, based for example on how close it is to the prediction. When the last pulse in this data segment has been examined, the data segment is divided into homogeneous segments and noise segments at 1122. Homogeneous segments are segments within which the blood circulation parameters, such as pulse shape, and systolic and diastolic pressure, are relatively uniform. If a parameter changes slowly during a long data segment, but cumulatively ends up changing significantly, then the data segment is arbitrarily divided up into shorter homogeneous segments, within each of which the parameters do not change very much. Data segments are also divided into homogeneous segments, not so arbitrarily, at points where parameters change suddenly.

Groups of consecutive pulses in which many or all pulses are labeled as noise pulses, are optionally set aside as noise segments. Such noise segments could be caused, for example, by the patient moving for several pulse beats, or by a sensor failing, or coming loose. Optionally, sporadic individual noise pulses that are surrounded by good pulses are still assigned to a homogeneous segment, although the parameters of those pulses, as least for the sensor data that is noisy, are optionally not included when finding average or typical values for the parameters for that homogeneous segment.

At 1124, typical parameters, including pulse shape, are found for each homogeneous segment, for example by averaging over the pulses in the segment, optionally excluding pulses that are outliers in some respect, for example pulses that were labeled as noise pulses at 1114. These typical parameters are reported at 1126, and, for example, are included in a chart such as the one shown below in FIG. 12.

If the segment has been identified at 1106 as a scaling segment, then the pulses are identified, and distinguished from large amplitude noise between pulses, at 1128. The most likely diastolic and systolic pressures, and their error bars, are then found at 1130, for example using the methods described in Figs. 6-8 for scaling events, or in Figs. 9 and 10 for mini-scaling events. This data is also reported at 1126. At 1132, the data is examined to see if there are any more segments, and if so, the next segment is examined at 1104. If there are no more segments in the data, the procedure ends at 1134.

FIG. 12 shows a chart 1200 with segmented data, produced for example by the method described in FIG. 11. Data 1202 from microphone 118 is plotted at the top, as a function of time, represented by a time axis 1204. The applied pressure 1206, from pressure sensor 112 for example, is plotted together with microphone data 1202. The first segment shown, at the left, is a scaling segment 1208, and within this segment the applied pressure first rises to a maximum, and then slowly falls to nearly zero. A systolic point 1214 and a diastolic point 1216 are identified in this data, using the methods described in Figs. 6-8 for example, and the corresponding systolic pressure 1218 and diastolic pressure 1220 are plotted at the bottom. Scaling segment 1208 is followed by a homogeneous segment 1210, during which microphone data 1202 has a very uniform envelope. Pressure data 1212, from force sensor 117 for example, is plotted at the bottom, and has an envelope in agreement with the systolic and diastolic pressures found in scaling segment 1208. Optionally, pressure data 1212 is scaled to match the systolic and diastolic pressures found in scaling segment 1208, since the scaling segment data may be more reliable than the absolute scale of the raw pressure data, which may depend on a contact area between force sensor 117 and the skin which may not be very accurately known. A typical pressure sensor pulse wave 1238, and a typical microphone pulse wave 1236, which may be proportional to the time derivative of pressure, for homogeneous segment 1210, are also plotted on chart 1200. Optionally, the microphone data 1202 and pressure data 1212 shown for homogeneous segment 1210 are not all of that sensor data for the segment, but only a small portion of it. The actual segment might last for much longer than the portion shown on chart 1200. Optionally, the portion of a homogeneous segment plotted is always a fixed fraction of its total length, for example 1% or 2% or 5%, so that a physician can see at a glance how long each segmented lasted.

Following segment 1210, there is another scaling segment 1222, with a new value 1224 of systolic pressure, and a new value 1226 of diastolic pressure. Following segment 1222 is a noise segment 1228, which has substantial noise in the microphone data plotted at the top. There follows another homogeneous segment 1230, another scaling segment 1232 with new values of systolic and diastolic pressure, and another homogenous segment 1234. Typical pulses 1240 and 1242 for segment 1230, and typical pulses 1244 and 1246 for homogeneous segment 1234, are also plotted on chart 1200.

FIG. 13 is a plot 1300 or microphone data 1302, and pressure sensor data 1304, with the same time axis 1306. Partly because the microphone signal is proportional to the time derivative of the pressure, the microphone is more sensitive to high frequency noise than the pressure sensor. A burst of noise 1308 in the microphone data, for example, is not visible in the corresponding pressure sensor data for interval 1310.

Because the microphone signal is proportional to the time derivative of the pressure, an independent measure of the pressure pulse wave shape may be found by integrating the microphone signal over time. FIG. 14 shows a plot 1400 of a time- integrated microphone signal 1402, and a pressure sensor signal 1404, for the same point on the same artery. The two signals look very similar. A pulse wave shape, more accurate than that given by either sensor alone, is optionally obtained by combining the integrated microphone signal and the pressure sensor signal. For example, in regions where there is high frequency noise, such as 1308 in FIG. 13, the microphone signal will generally be more sensitive to the noise than the pressure sensor, and the pressure sensor signal will be given more weight than the integrated microphone signal. But in other regions, the integrated microphone signal is optionally given more weight than the pressure sensor signal, because the microphone signal may more accurately reflect the higher frequency components of the pulse wave. For example, if the pressure sensor is a load cell, it may have very low sensitivity above about 10 Hz, and even the time derivative of the pressure sensor signal may have low sensitivity above 10 Hz, while the microphone may be most sensitive above 10 Hz, and have substantial sensitivity even at hundreds or thousands of Hz.

Because the time-integrated microphone signal has an unknown integration constant, the pressure sensor signal is used to determine the absolute value of the pressure, as opposed to differences in pressure, in the pulse wave shape. In addition, the values of the diastolic and systolic pressure are corrected, when they become too uncertain, by initiating a scaling or mini-scaling event. It should also be noted that there are regular changes in blood pressure associated with 1) breathing, 2) a regular cycle approximately 90 minutes long, and 3) a diurnal effect, associated with temperature. When using scaling or mini- scaling events, more accurate results can be obtained by taking into account at what phase in these cycles the systolic and diastolic pressures were measured.

The pulse wave shape may be analyzed to obtain clinically useful information. Integrated microphone signal 1402, proportional to the pressure, shows a minimum 1408, corresponding to the diastolic pressure. There is a rapid rise 1410, ending at a maximum 1412 corresponding to the systolic pressure. The maximum rate of rise of the pressure is used as an index of cardiac contractility. After peak 1412, the pressure falls more slowly than it rose. It reaches a local minimum 1414, called the dichrotic notch, followed by a second local maximum 1416, which is the peak of the reflected wave, due to the impedance mismatch between the larger arteries and the capillaries. The time delay between peak 1412 and reflected peak 1416, and the depth of the dichrotic notch, are measures of arterial elasticity and stiffness. Children, with very elastic arteries, have a large time delay between peak 1412 and reflected peak 1416, and a deep dichrotic notch. Elderly people, with very stiff arteries, have a shallow dichrotic notch or none at all, with reflected peak 1416 not well separated from peak 1412. Following reflected peak 1416, the pressure falls more slowly until it reaches the next minimum 1418 at the diastolic pressure. Other clinically useful parameters that may be derived from the pulse wave shape will be described below in FIG. 18. Using a combination of the pressure sensor data and time-integrated microphone data gives a pulse wave shape that is potentially more accurate than the pulse wave shape obtained from a pressure sensor alone, and may be useful clinically. In particular, it may allow a relatively accurate method to measure of the pulse shape that, unlike a direct A-line measure of pulse wave shape, is not invasive.

FIG. 15 illustrates an objective way to define different parts of a pulse wave, which are needed for calculating some blood circulation parameters. Plot 1510 shows the pulse wave pressure as a function of time. Plot 1520 shows the second derivative of plot 1510, and plot 1530 shows the fourth derivative of plot 1510. Time 1502, the end of diastole and the beginning of systole, may be defined as the first peak in the second derivative shown in plot 1520. Time 1504, the end of systole and beginning of diastole, may be defined as the second local peak in the second derivative, following the peak pressure. Time 1506, the time of the reflected peak, may be defined as the third zero of the fourth derivative after the beginning of systole, which goes from negative to positive. Alternatively, the time of the reflected peak is defined as the last zero of the fourth derivative going from negative to positive, before the end of the systole. This alternative definition may be more robust if, as in plot 1530, there is marginally a pair of zeroes before time 1506. These definitions may be useful even in cases, such as the aortic pulse wave, where there may be no local maximum in pressure at the reflected peak.

FIG. 16 is a flowchart 1600 for a method of measuring pulse wave velocity. Pulse wave velocity is useful for tracking short term changes in blood pressure, in between scaling events, since the pulse wave velocity depends on blood pressure, for a given arterial elastance, which does not generally change over the short term. The higher the blood pressure, the faster the pulse wave velocity, for a given value of arterial elastance.

At 1602, two sensors are placed along an artery, displaced from each other by a short distance, for example less than 10 cm, or less than 5 cm, or about 2.5 cm. The two sensors can be, for example, vibration sensor 116 and microphone 118 in FIG. 1. Optionally, the two sensors are the same type of sensor, and even identical models. Optionally, they are both microphones, with output proportional to the time derivative of the pressure. The pulse wave velocity typically ranges between 5 m/sec and 15 m/sec, with the lower velocity typical of children, and the higher velocity typical of elderly people with inelastic arteries. If two sensors are about 2.5 cm apart, as they are in the case of sensors 116 and 118 in FIG. 1, then the pulse wave delay (pulse travel time) between the pulse wave at the two sensors is between 1.7 and 5 milliseconds. This is comparable to the sampling time for such sensors, so it is difficult to obtain accurate measurements of pulse wave velocity by directly comparing the pulse wave shape obtained by the two sensors. In the prior art, pulse wave velocity has generally been measured by sensors a meter apart or more, for example one sensor on the chest, near the heart, and the other sensor on the wrist, or on the ankle. Instead, the method outlined in flowchart 1600 makes use of a comparator pin on each of the microprocessors running the sensors. This pin provides the time, to within one microprocessor clock cycle, when the output signal of the sensor crosses zero volts, or any other voltage, in a particular direction, for example negative to positive. If the microprocessor has a clock rate of 20 MHz, then the time provided by the microprocessor is accurate to within 20 nanoseconds, far shorter than the pulse travel time between the two sensors. At 1604, the crossing voltage is chosen, for example zero, and the comparators are set to provide the crossing time. At 1606, the difference between the crossing times is recorded, for one or more pulses, to obtain the pulse wave delay between the two sensors. At 1608, the pulse wave delay is optionally averaged over many pulse waves. At 1610, the displacement between the two sensors is divided by the (optionally averaged) pulse wave delay, to obtain the pulse wave velocity. A potential advantage of using this method, over using sensors located at the chest and on the wrist or ankle, is that the entire system can be included in a single unit, worn for example on the wrist, which is easier to use. FIG. 17 is a plot 1700 of pulse wave signals from two microphones, displaced a few cm apart along an artery, illustrating the phase delay due to the pulse travel time. The microphone closer to the heart has a signal 1702 plotted as a solid line, and the microphone further from the heart has a signal 1704 plotted as a dashed line. In general, it is difficult to compare the phase of the two signals very precisely, because they have somewhat different amplitudes, due to different coupling of the microphones to the artery. However, if the times when the signal crosses zero is used for comparison, the phase delay between the two signals is fairly reliable, because the two zero crossings in each pulse cycle correspond to the times of diastolic and systolic pressure. A crossing of zero from negative to positive, corresponding to the diastolic point, occurs at times 1706, 1710, and 1714 for signal 1702, and at times 1708, 1712, and 1716, for signal 1704. The phase delay time is close to 5 milliseconds on average, with approximately a 25% variation from pulse to pulse. A crossing of zero from positive to negative, corresponding to the systolic point, occurs at times 1718, 1722, and 1726 for signal 1702, and at times 1720, 1724, and 1728 for signal 1704. In this case, the phase delay time is also close to 5 milliseconds, with less variation from pulse to pulse. For this reason the phase delay time for the systolic point may be a better choice than the phase delay time for diastolic point. Alternatively, both phase delays may be measured and used, averaging over many pulses, to find the pulse travel time.

FIG. 18 is a flowchart 1800 showing a procedure for calculating various blood circulation parameters, that may be clinically useful, from the pulse wave shape, pulse wave velocity, and other parameters that can be measured non-invasively using a system such as system 100 in FIG. 1. All of the steps shown in flowchart 1800 are optional, depending on which parameters are to be calculated. At 1802, the pulse wave shape is found optionally in the radial artery, or alternatively in another peripheral artery such as the tibial artery. Optionally, the pulse wave shape is found using a combination of a time-integrated microphone signal and a pressure sensor signal, as shown in FIG. 14. Optionally other sensors are used as well. At 1804, a transfer function is used to find the aortic pulse wave shape from the radial (or other peripheral) artery pulse wave shape, as described for example in US published patent application 2004/0199080 to Tanabe, and in K. Takazawa et al, Hypertension Research 30(3), 219-228 (2007).

At 1806, the start and end times of the systole and diastole are found, as described in FIG. 15. At 1808, the systolic inflection point is found for the radial artery and aorta pulse wave. This is the point at time 1506, defined in terms of the fourth derivative of the pressure, in FIG. 15, which corresponds to the peak of the reflected wave in the radial artery. At 1810, the time of the aortic valve incisures is found, using the signal from microphone 118. At 1812, the aortic and radial artery systolic integral percentage is calculated, from the integral of the pulse wave between the start and end of the systole and the total integral of the pulse wave for one pulse, as shown in FIG. 15. At 1814, the augmentation index, a measure of the relative amplitude of the forward and reflected pulse wave, is found for the aortic and radial artery pulse waves. It is defined, for example, as the ratio of the height of the reflected peak pressure above the diastolic pressure, to the ratio of the height of the systolic peak pressure above the diastolic pressure, and it is an indication of systemic vascular resistance (SVR). In addition to being a measure of age-related changes in arterial stiffness, SVR is higher immediately after by-pass surgery, and may be monitored during the first few hours of recovery from by-pass surgery.

At 1816, the maximum rate of rise of the pulse wave is found, and the cardiac contractility. Cardiac output is also correlated with the rise time, and may be estimated. Generally, cardiac output per heart beat is inversely proportional to the rise time of the pulse wave. However, corrections may have to be made for the differences in pulse wave near the heart (in the aorta or carotid artery) and in the radial artery, using "End Pressure" theory in fluid dynamics, as described, for example, in www. mi- labs. co.jp/RD2-e.htm, cited above. At 1820, the pulse wave velocity is found, for example using the method shown in Figs. 16 and 17. At 1822, the systemic vascular resistance, is found, using the augmentation index found in 1814. At 1824, the effective arterial elastance index and arterial compliance are found, using, for example, the formulas given by Kelly, and by Liu et al, cited above. At 1826, the aortic mean pressure is found, from the aortic pulse wave shape. At 1828, the aortic-radial systolic pressure gradient (difference between aortic and radial artery systolic pressure) is found.

As used herein the term "about" refers to ± 10 %.

The terms "comprises", "comprising", "includes", "including", "having" and their conjugates mean "including but not limited to". This term encompasses the terms "consisting of" and "consisting essentially of".

The phrase "consisting essentially of" means that the composition or method may include additional ingredients and/or steps, but only if the additional ingredients and/or steps do not materially alter the basic and novel characteristics of the claimed composition or method. As used herein, the singular form "a", "an" and "the" include plural references unless the context clearly dictates otherwise. For example, the term "a compound" or "at least one compound" may include a plurality of compounds, including mixtures thereof.

Throughout this application, various embodiments of this invention may be presented in a range format. It should be understood that the description in range format is merely for convenience and brevity and should not be construed as an inflexible limitation on the scope of the invention. Accordingly, the description of a range should be considered to have specifically disclosed all the possible subranges as well as individual numerical values within that range. For example, description of a range such as from 1 to 6 should be considered to have specifically disclosed subranges such as from 1 to 3, from 1 to 4, from 1 to 5, from 2 to 4, from 2 to 6, from 3 to 6 etc., as well as individual numbers within that range, for example, 1, 2, 3, 4, 5, and 6. This applies regardless of the breadth of the range.

Whenever a numerical range is indicated herein, it is meant to include any cited numeral (fractional or integral) within the indicated range. The phrases "ranging/ranges between" a first indicate number and a second indicate number and "ranging/ranges from" a first indicate number "to" a second indicate number are used herein interchangeably and are meant to include the first and second indicated numbers and all the fractional and integral numerals therebetween.

As used herein the term "method" refers to manners, means, techniques and procedures for accomplishing a given task including, but not limited to, those manners, means, techniques and procedures either known to, or readily developed from known manners, means, techniques and procedures by practitioners of the chemical, pharmacological, biological, biochemical and medical arts.

It is appreciated that certain features of the invention, which are, for clarity, described in the context of separate embodiments, may also be provided in combination in a single embodiment. Conversely, various features of the invention, which are, for brevity, described in the context of a single embodiment, may also be provided separately or in any suitable subcombination or as suitable in any other described embodiment of the invention. Certain features described in the context of various embodiments are not to be considered essential features of those embodiments, unless the embodiment is inoperative without those elements. Various embodiments and aspects of the present invention as delineated hereinabove and as claimed in the claims section below find experimental support in the following examples.

EXAMPLES

Reference is now made to the following examples, which together with the above descriptions, illustrate some embodiments of the invention in a non limiting fashion.

FIG. 19 shows data for a scaling event comprising a pressure signal 710, a first derivative 715 of the pressure signal, a microphone signal 720, and a load-cell signal 725, all plotted as functions of time. Probability functions 750, 760, and 770 are found for the systolic blood pressure to occur at each pulse beat, based separately on the first derivative of pressure, a low frequency part of the microphone signal, and an AC component of the load-cell signal respectively. These probability functions are not probabilities, since they are not normalized to 1, but are designed to be higher for pulse beats that are more likely to correspond to the systolic pressure, and lower for those pulse beats that are less likely to correspond to the systolic pressure. Probability functions 750, 760 and 770 are then combined, for example they are multiplied together, to give an overall probability function 740 for the systolic pressure to occur at each pulse beat. Optionally one or more other data signals such as a detrended pressure signal, a high frequency part of the microphone signal, a DC load cell signal, PWV, and/or oximetry are also used to find probability functions of systolic pressure, and also contribute to overall probability function 740. The pressure 710 at the time of greatest overall probability for the systolic pressure is then optionally identified as the most likely value for the systolic pressure, which is 110 mm Hg in the case shown in FIG. 19. Optionally, the width of probability function 740 provides approximate error bars or confidence limits on the most likely values for systolic pressure. A similar procedure is optionally used with some or all of the same data to find an overall probability function and most likely value for diastolic pressure.

In some embodiments of the invention, instead of treating all of the probability functions equally when multiplying them together to obtain an overall probability function, some of them are given more weight than others. This is done, for example, by defining a total score function Score(i) for the systolic pressure, in terms of probability functions P,(i) and corresponding weights W_j, for each pulse beat i:

Score(i) = ∑W_] ln(P_] (i))

An overall probability function, similar to function 740, may then be defined as the exponential of Score(i), but the overall probability function need not be calculated, and Score(i) may be used directly to find a most likely value of the systolic pressure and its approximate error bars. A similar score function may be defined for the diastolic pressure. Optionally, there is more than one probability function P_j for one or more of the data signals, which is similar to defining a single probability function which is a product of the different probability functions, for that data signal. Optionally, the weights W_j are selected empirically, to give a good fit to the systolic pressure as measured by other means, for example by invasive or non-invasive means, e.g. prior art methods. Optionally, all of the weights W, are set equal to 1. In some embodiments of the invention, the functions P_j are combined in a different way to obtain a total score function.

In an exemplary embodiment of the invention, probability functions 750, 760 and 770 are based on specific characteristics of the data signals, as will be explained below. An a priori probability function may also contribute to the total score function, based for example on a distribution of past values of systolic pressure measured for the same patient, optionally weighted by how recent the measurement was. For example, the a priori probability function may be a Gaussian distribution of the applied pressure at the time of pulse beat being considered, with the same mean and standard deviation as a distribution of past values of systolic pressure.

Although the procedure is described herein as involving probability functions, it may also be understood in terms of fuzzy logic, with the probability functions interpreted as fuzzy membership values.

Optionally, a probability function over pulse beats is calculated for each data signal, based only on features of that data signal for that pulse beat and optionally for neighboring pulse beats, independent of the other data signals, and this was done for probability functions 750, 760, and 770 plotted in FIG. 19. Alternatively, the probability functions over pulse beats for one or more data signals depend also on features of other data signals for that pulse beat and optionally for neighboring pulse beats. In some embodiments of the invention, the probability functions for one or more data signals may depend only features of the data for that pulse beat and for past pulse beats, and the probability function may be evaluated in real time. In other embodiments of the invention, however, the probability functions depend also on the data at future pulse beats, and are only evaluated after all of the data has been received, or after all of the data has been received that is needed for the evaluation. This is the case for the probability distributions plotted in FIG. 19, for example, which depend on the value and timing of a global maximum of each data signal.

Optionally, the probability function for the systolic pressure, using the data signal that is the first time derivative of the pressure, has its greatest value when the peak first derivative of pressure, for a given pulse beat, is closest to 0.6 times the global maximum peak first derivative of pressure for any pulse beat. The probability function is optionally set to zero for pulse beats that occur after the pulse beat with the global maximum peak first derivative. Optionally, for example to avoid having to take the logarithm of zero when calculating the total score function Score(i), the probability function is set to an arbitrary small number, such as e^"10, which will ensure that Score(i) does not have its maximum value at any of those pulse beats. For pulse beats occurring before the global maximum, the probability is optionally a Gaussian function of the peak first derivative of the pressure, with the peak of the Gaussian occurring at a peak first derivative equal to 0.6 times the global maximum peak first derivative, and the standard deviation chosen empirically. For example, for probability distribution 750 plotted in FIG. 19, the standard deviation of the Gaussian is 0.2 times the global maximum first derivative. In this example, the probability function is two standard deviations away from its highest value, and hence quite small, at the pulse beat where the peak first derivative reaches its global maximum, and at early pulse beats where the peak first derivative is only 20% or less of the global maximum, where the data may be dominated by noise.

Calling the peak value of this data signal S₁, for pulse beat i, this expression for the probability function for this data signal may be written: t ^Tι < ^~ t 'max

Here ^ means the time of pulse beat i, t_max means the time of the maximum peak data signal s_max, and ε indicates an arbitrary small number such as e^~10. The subscript "FDP" indicates the data signal "first derivative of pressure" and the superscript "(sys)" indicates that this probability function is for the systolic pressure. The standard deviation σ is, for example, set equal to 0.2s_max, where s_max is the maximum S₁ for any pulse beat i in the data set for this signal. Alternatively, instead of using s_max, the peak values S₁ as a function of, for example, time, or as a function of pulse beat i, are fitted to a curve, for example using cubic splines or any other known curve-fitting method, and the peak value of the curve is used. This peak value is generally similar to s_max, but may be less sensitive to the exact timing of the pulse with respect to the applied pressure. The inventors believe that this peak value may occur when the applied pressure is close to the mean arterial pressure.

A similar expression is optionally used for the probability function P_FDP ^(dia)(i) for the diastolic pressure, using the data signal that is the first time derivative of the pressure, but with the probability distribution set at zero (or an arbitrary small value) for U < Wax, and with the peak of the Gaussian occurring (for U > t_max) when S₁ = 0.8s_max. The standard deviation σ in this case is optionally set equal to 0.1s_max, so the probability function will be very small, two standard deviations from its highest value, when S₁ is equal to s_max. For some data signals, such as the load cell signal, no probability function has been clearly identified so far as useful for finding the diastolic pressure, and optionally those data signals are not used when finding a total score function for the diastolic pressure.

In some embodiments of the invention, instead of or in addition to using the first derivative of the pressure as one of the data signals, a de-trended (AC) pressure signal is used as a data signal. The de-trended pressure is found, for example, by fitting the pressure signal to a straight line or an exponential, subtracting the straight line or exponential, and optionally adjusting the remaining signal so that its minimum is zero for each pulse beat. The probability functions Pop^(sys)(i) and Pop^(dia)(i) for the systolic and diastolic pressures (where "DP" means "de-trended pressure") optionally depend on the peak de-trended pressure signal S₁ for each pulse beat, and on the global maximum S_max of the peak de-trended pressure signals for any pulse beat. The probability functions are then optionally defined similarly to the case where the first derivative of the pressure is used as the data signal.

For the microphone data signal, a similar expression is optionally or alternatively used for the probability function of the diastolic pressure, but optionally with the peak of the Gaussian occurring at the pulse beat i where the peak signal S₁ for that pulse beat is equal to the global maximum peak signal s_max. A different kind of expression is optionally used for the probability function for the systolic pressure, for the microphone signal and the AC load cell signal, which depends on the peak signals having an upward trend from one pulse beat to the next, as will be described below. Such expressions are used for probability functions 760 and 770 plotted in FIG. 19.

The probability function over pulse beats for the systolic pressure, for the microphone and load cell data signals, is optionally the product of two expressions. Since the different probability functions are multiplied together when finding the total score or overall probability function, each of these expressions may be treated as a separate probability function for that data signal, which will then have more than one probability function defined for it.

In an exemplary embodiment of the invention, the first probability function is designed to be large if the neighboring pulse beats to the pulse beat being examined show a monotonic upward trend in the peak of the data signals S₁ for successive pulse beats i, and to be small if the neighboring pulse beats do not show an upward trend. This probability function, which will be referred to as P_mOno(i), where "mono" indicates "monotonic," is optionally given by

The first expression in parentheses depends on the peak value of the signal for the next r-1 future pulse beats, and is large if the peak values of the signal for these future pulse beats are large. The second expression in parentheses depends on the peak value of the signal for the previous r pulse beats, and is large if the peak values of the signal for these past pulse beats are small. Thus, P_mono can be greatest if the peak signal is larger for the future pulse beats and smaller for past pulse beats, which will be the case if it is monotonically increasing. The inventors have found that r = 4 works well, but alternatively r may be any small positive integer, such as 1, 2, 3, 5, 6, 10, or a larger or intermediate value.

In some embodiments of the invention, the number of past pulse beats examined is not necessarily one more than the number of future pulse beats examined, in the expression for P_mono- For example, P_mono could depend only on comparing present to future pulse beats, or only on comparing present to past pulse beats. In that case, only

o^r only would appear in the expression for P_mono, and the other

one would optionally be replaced by S₁ to keep P_mono dimensionless. Similarly, the number of values of / that each of these products runs over could be different for the two products, with the missing si 's replaced by S₁ 's, to keep P_mono dimensionless. In any case, this is optionally done for pulse beats i near the beginning of the data set, or pulse beats i near the end of the data set, where the full range of pulse beats / in the product is not available. In that case, the missing si 's are optionally replaced by S₁ 's.

Optionally, the second probability function is designed to be greatest when the peak of the data signal for the pulse beat being considered has a particular amplitude, much less than the global maximum peak value for any pulse beat, but not so low that it is likely to be dominated by noise. This probability function P_amp(i), where the subscript "amp" indicates "amplitude," is used in order to select a value for the systolic pressure that occurs when the peak microphone or load cell data signal for pulse beat i has an amplitude that is small, but clearly rising out of the noise. For example,

Here, θ and k are free parameters, and F is a gamma function, used to normalize the expression. The maximum value of P_amp(i) occurs when S₁ = (k - l)θ. The parameters θ and k are optionally chosen empirically, so that the probability distribution gives good results in picking out the systolic pressure. The inventors have found that θ = 0.25s_max, and k = 2.2, which gives the maximum P_amp(i) at S₁ = 0.3s_max, works well. Other values of k and θ, for which (k - l)θ is well below s_max, but not so low that it is comparable to the noise level of the data signal, may also work well.

Probability functions 760 and 770, as plotted in FIG. 19, are respectively Pmono(i)P_amp(i) for the low frequency component of the microphone data signal, and the AC component of the load cell data signal, calculated using the expressions given above.

In some embodiments of the invention, additional probability functions are also used for the low frequency microphone data signal and the AC load cell data signal, for the systolic pressure. One such probability function selects for peaks in the data signal that correspond to real pulse beats, and not just to noise. This function is greatest when the time interval from the pulse beat being considered to the next pulse beat is within the range of typical time intervals between pulse beats, based on the late time part of the data signal, when the applied pressure is well below the systolic pressure, and the timing of the pulse beats is clear. The inventors have found that this probability function, designed P_Δt(i), may be useful for picking out the first pulse beat after the systolic point, which is often only slightly above the noise level. This probability function is defined as:

Here At₁ = t₁₊i - U is the time interval between pulse beat i (or rather, the peak i in the data signal that is believed to be a pulse beat) and the following pulse beat. The mean interval between pulse beats, Δt_mean, and the standard deviation σ of the intervals between pulse beats, are optionally evaluated using pulse beats relatively late in the data signal, at applied pressures that are well below the systolic pressure, where the peaks in the data signal corresponding to pulse beats are well above the noise and easy to identify. Optionally, if t₁₊₂ - U is closer to Δt_mean than t₁₊i - U is, then At₁ is defined as t₁₊₂ - U instead of t₁₊i - t_l5 or At₁ is defined as t_1+n - U for whatever positive integer n makes this quantity closest to Δt_mean. This modification in the definition of At₁ has the potential advantage that peaks that are due to noise, rather than to real pulse beats, may be less likely to affect the value of P_Δt(i). The inventors have found that choosing K = 4 works well for distinguishing early pulse beats in the microphone or AC load cell data signals, after the applied pressure has dropped below the systolic pressure, from noise in these data signals before the applied pressure has dropped below the systolic pressure. Other values of K, smaller or larger, may also be useful. Another probability function that may be useful to use with the low frequency microphone signal and the AC load signal, for example, for finding the systolic pressure, is a probability function Pp_2B(I), where "P2B" indicates "peak to baseline ratio." This function is the ratio of the height of the peak of the data signal, at pulse beat i, above a mean value of a baseline signal, to the standard deviation of the baseline signal. Here the baseline signal for pulse beat i is optionally defined as the signal in that part of the interval between the peak of the previous pulse beat, and the peak of pulse beat i, where the signal is lower than the mean value of the signal between the peak of the previous pulse beat and the peak of pulse beat i. If the peak being considered is at the first pulse beat after the systolic point, then the previous peak is likely to be due to noise, rather than to a pulse beat, and will have much lower amplitude, so the standard deviation of the baseline signal in that interval will be at the noise level, much lower than the height of the signal at the peak being considered. The first pulse beat i after the systolic point is likely to have a much higher value of Pp2β(i) than earlier pulse beats, where the signal is mostly or entirely noise, and a somewhat higher value than later pulse beats.

The inventors have also found probability functions that may be useful for other data signals. One such signal is a high frequency component of the microphone signal. In the tests done by the inventors, the high frequency part of the microphone signal, well above the pulse frequency, was dominated by two frequency ranges close to 18 Hz and 36 Hz. The microphone signal was filtered to keep only the frequencies near 36 Hz. For different subjects and different equipment, it is possible that there will be different dominant frequency ranges in the high frequency part of the microphone data signal, and if so the signal is optionally filtered to include one or more of those frequency ranges instead. After filtering the microphone signal, the absolute value of the signal is taken, and filtered with a low pass filter. This produces a signal similar to the envelope of the high frequency signal. Alternatively, the envelope of the high frequency signal is taken, or any other procedure is used which produces a similar processed signal.

It has been found that a probability function similar to PΔt(i) is useful to use with the high frequency microphone signal for the systolic pressure. However, unlike the low frequency microphone signal, the high frequency microphone signal often does not have clearly defined peaks at pulse beats, when the applied pressure is well below the systolic pressure. Optionally, instead of obtaining the mean time interval Δt_mean and the standard deviation σ from the peaks in the data at low applied pressure, the mean time interval and standard deviation are found from all of the peaks in the data. Optionally, for example in order to reject outlying time intervals that may be due to peaks down in the noise, the time intervals between adjacent peaks are binned, the most populated bins are fitted to a Gaussian function, and the mean and standard deviation of the fitted Gaussian function are used when calculating PΔt(i), using the formula given above. Another probability function, that has been found to be useful for the high frequency microphone data signal, for the systolic pressure, is the ratio of the amplitude of the peak at the pulse beat being considered, to the amplitude of the previous peak. For the first pulse beat after the systolic point, the previous peak is likely to be due to noise, and to have a much lower amplitude, so this probability function is likely to have a higher value for this first pulse beat than for later and earlier pulse beats.

A DC (or low frequency) component of the load cell data has also been found to be useful for finding the systolic pressure and the diastolic pressure. This signal is optionally found by applying a low pass filter to the load cell data signal. Because the load cell measures the pressure in the artery downstream from where it is occluded, the signal tends to have a first local minimum at a time T_1111nI near when the applied pressure is equal to the systolic pressure, have an increasing value as the applied pressure is reduced, and a maximum near a time T_max when the applied pressure is equal to the diastolic pressure. A possibly useful probability function for this signal, for the systolic pressure, is a Gaussian function of the difference between the time ^ of pulse beat i, and T_nUn1, with a standard deviation that is optionally several times smaller than T_max - T_mini, for example 5%, 10%, or 20% of T_max - T_1111nI. A possibly useful probability function for this signal, for the diastolic pressure, is a Gaussian function of the difference between the time ^ of pulse beat i, and T_max, with the same standard deviation. Other values of the standard deviation are optionally used, and they need not be the same for the probability functions used for finding the systolic pressure and the diastolic pressure.

Although the invention has been described in conjunction with specific embodiments thereof, it is evident that many alternatives, modifications and variations will be apparent to those skilled in the art. Accordingly, it is intended to embrace all such alternatives, modifications and variations that fall within the spirit and broad scope of the appended claims.

All publications, patents and patent applications mentioned in this specification are herein incorporated in their entirety by reference into the specification, to the same extent as if each individual publication, patent or patent application was specifically and individually indicated to be incorporated herein by reference. In addition, citation or identification of any reference in this application shall not be construed as an admission that such reference is available as prior art to the present invention. To the extent that section headings are used, they should not be construed as necessarily limiting.

Claims

WHAT IS CLAIMED IS:

1. A system for measuring blood circulation parameters in a patient, comprising: a) a pressing element; b) a contact microphone which is acoustically coupled to the patient' s body by the pressing element at a location adjacent to an artery, the microphone thereby producing a data signal indicative of blood circulation in the artery; and c) a controller which integrates in time the data signal from the microphone, thereby determining a blood pressure as a function of time, at least up to an unknown integration constant.

2. A system according to claim 1, also including a pressure sensor which is coupled to the patient's body at the same or a different location adjacent to said artery, thereby producing a data signal indicative of a blood circulation in the artery, wherein the controller uses the data signal from the pressure sensor to determine said integration constant.

3. A system according to claim 2, wherein the controller combines the integrated data signal from the microphone and the data signal from the pressure sensor to find a most probable blood pressure as a function of time.

4. A system according to claim 1, wherein the microphone is capable of detecting, above a noise level, rates of change of pressure of at least 20 mm Hg per second, for frequencies at least as low as 1 Hz, in no more than 0.1 seconds.

5. A system according to claim 1, wherein the pressing element acoustically couples the microphone to the patient's body at a location adjacent to the radial artery, the brachial artery, the posterior tibial artery, the femoral artery, or the carotid artery.

6. A system for measuring a pulse wave velocity in an artery of a patient, comprising: a) a first and a second sensor of arterial pulse data, in a positioning element which couples the sensors to the body of the patient adjacent to an artery, spaced at a known separation distance of less than 10 cm along the artery, each sensor associated with an output channel conveying a data signal and with a comparator that provides an indication of a time at which the data signal had a specific value; and b) a controller which measures a pulse wave delay time between the first and second sensor, by comparing time indicated by the comparators of the first and second sensors.

7. A system according to claim 6, wherein the known separation distance is less than 5 cm.

8. A system according to claim 7, wherein the known separation distance is less than 3 cm.

9. A system according to claim 6, wherein the two sensors measure a same kind of data.

10. A system according to claim 9, wherein the two sensors are substantially identical.

11. A system according to claim 6, wherein the two sensors comprise one or both of a microphone and a pressure sensor.

12. A system according to claim 6, wherein the specific value corresponds to systolic time for both sensors, or corresponds to diastolic time for both sensors.

13. A system according to claim 6, wherein the indication of time for each sensor comprises an identification of a clock cycle of the comparator for that sensor, and the clock cycle is shorter than 0.1 millisecond.

14. A system for estimating systolic blood pressure of a patient, comprising: a) a pressure applicator which applies a controllable pressure on an artery of the patient; b) a sensor which measures an amplitude of a pulse beat in the artery when the pressure applicator is applying the controlled pressure; and c) a controller which: i) varies the applied pressure over a range of pressures low enough not to occlude blood flow in the artery completely, over a period covering a plurality of pulse beats, while receiving data from the sensor; ii) determines the pulse amplitude as a function of the applied pressure; and iii) estimates the systolic pressure by extrapolating the data to an applied pressure at which the amplitude of the pulse beat would go to zero.

15. A system according to claim 14, wherein the controller varies the applied pressure by increasing it to value at which the pulse amplitude is reduced by a factor of at least two from its value at the lowest applied pressure in the range.

16. A system according to claim 15, wherein the factor is at least five.

17. A system according to claim 14, wherein the lowest applied pressure in the range is less than an expected diastolic pressure of the patient, and the controller also estimates the diastolic pressure from the pulse amplitude as a function of applied pressure.

18. A system according to claim 14, wherein the sensor is a contact microphone.

19. A system according to claim 14, wherein the sensor is a pressure sensor.

20. A method for determining a blood pressure as a function of time, at least up to an unknown integration constant, for a patient's artery, comprising integrating in time a data signal indicative of a pulse beat in the artery, obtained from a contact microphone acoustically coupled to the patient's body adjacent to the artery.

21. A method according to claim 20, wherein the data signal is integrated for at least one pulse cycle, thereby obtaining a pulse wave shape of the pressure as a function of time.

22. A method according to claim 20, also including acoustically coupling the microphone to the patient's body adjacent to the artery, and generating the data signal indicative of the pulse beat in the artery.

23. A method for measuring a pulse wave velocity in a patient's artery, comprising: a) receiving from each of two comparators in microprocessors associated with two sensors coupled to the patient's body adjacent to the artery and separated by a known separation distance along the artery, an indication of a time when a data signal from that sensor had a specific value; b) finding a pulse wave delay time from a difference between the times indicated for the two sensors; and c) calculating the pulse wave velocity from the separation distance and the pulse wave delay time.

24. A method according to claim 23, also including coupling the two sensors to the patient's body adjacent to the artery and separated by the separation distance, and generating the data signal from each sensor.

25. A method of estimating systolic blood pressure of a patient, comprising: a) applying a pressure varying over a range to an artery of the patient, over a period of a plurality of pulse cycles and substantially constant during each pulse cycle, with the maximum pressure in the range partially occluding blood flow in the artery but not completely occluding the blood flow; b) generating data on a pulse amplitude in the artery while applying the varying pressure; c) determining the pulse amplitude as a function of the applied pressure; and d) extrapolating the data to a pressure at which the pulse amplitude would go to zero, thereby estimating the systolic pressure.

26. A method according to claim 25, wherein applying the varying pressure comprises increasing the pressure from the minimum pressure in the range to the maximum pressure in the range.

27. A method according to claim 26, comprising stopping the increasing of the varying pressure, responsive to the data on pulse amplitude.

28. A method according to claim 25, also including: e) repeating (a) through (d) one or more times; f) recording the estimated systolic pressure each time (d) is repeated; and g) finding an average of the recorded estimated systolic pressures, to obtain a more accurate estimate of the systolic pressure.

29. A system for measuring a breathing parameter of a patient, the system comprising: a) a set of one or more sensors suitable for measuring beat-to-beat blood pressure in an artery of the patient; and b) a controller which uses data from the sensors to detect beat-to-beat changes in one or more blood pressure parameters indicative of the patient's breathing cycle, thereby measuring the breathing parameter.

30. A system according to claim 29, wherein the controller uses the data to measure a breathing parameter indicative of a sleep disorder.

31. A method of measuring a breathing parameter in a patient, the method comprising: a) measuring a blood pressure parameter in the patient for an interval comprising a plurality of breathing cycles; b) analyzing changes in the blood pressure parameter from beat to beat over a breathing cycle timescale; and c) extracting the breathing parameter from the changes in the blood pressure parameter.

32. A method according to claim 31, also including using the breathing parameter to diagnose a sleep disorder.

33. A method of obtaining homogeneous segments of data on blood circulatory parameters, comprising: a) generating data of at least one circulatory parameter as a function of time; b) identifying as homogeneous segments one or more time intervals, within each of which said parameter is relatively homogeneous; and c) calculating a representative sample of the data for each homogeneous segment.

34. A method according to claim 33, whereby calculating a representative sample comprises choosing a representative subset of the data in said homogeneous segment.

35. A method according to claim 33, whereby calculating a representative sample comprises finding an average over said homogeneous segment.

36. A method according to claim 33, also comprising identifying scaling segments of data from scaling events.

37. A method according to claim 33, also comprising identifying as noise segments, one or more time intervals within which the data has atypical values, and excluding the noise segments from consideration as homogeneous segments.

38. A system for determining a set of one or more blood circulation parameters, comprising: a) one or more sensors for obtaining a plurality of data signals pertaining to the blood circulation parameters; and b) a controller which, for each parameter, calculates a probability distribution for the value of the parameter from each data signal, and calculates an overall probability distribution for the value of the parameter by combining the probability distributions for each data signal.

39. A system according to claim 38, also including a pressure applying element for applying a variable known pressure to an artery, wherein the data signals comprise pulse data obtained from the artery while the variable known pressure is applied to the artery, and the blood circulation parameters comprise one or both of systolic pressure and diastolic pressure.

40. A system according to claim 38, wherein the sensors comprise a microphone and a pressure sensor, the data signals comprise a first signal from the microphone and a signal from the pressure sensor, and the blood pressure parameters comprise information on pulse wave shape.

41. A system according to claim 40, wherein the data signals also comprise a second signal from the microphone, the first signal being filtered to relatively increase lower frequencies, and the second signal being filtered to relatively increase higher frequencies.

42. A method for determining a set of one or more blood circulation parameters, comprising: a) obtaining a plurality of data signals pertaining to the blood circulation parameters; b) calculating, for each parameter, a probability distribution for the value of the parameter from each data signal; and c) calculating an overall probability distribution for the value of the parameter by combining the probability distributions for each data signal.

43. A method according to claim 42, wherein the data signals comprise pulse data obtained from an artery while a variable known pressure is applied to the artery, and the blood circulation parameters comprise one or both of systolic pressure and diastolic pressure.

44. A method according to claim 43, wherein calculating a probability distribution, for one or both of the systolic and diastolic pressure, comprises: a) calculating from the data signal a probability distribution of the time of occurrence of said category of pressure; b) finding the value of the variable known pressure as a function of time for a range of time of said distribution of the time of occurrence; c) obtaining a measure of stiffness of the artery; and d) using the measure of stiffness of the artery to correct said value of the variable known pressure as a function of time for the range of time, to obtain said probability distribution for said category of pressure.

45. A method according to claim 44, wherein obtaining the measure of stiffness of the artery comprises obtaining a measure of stiffness based on a measurement of pulse wave velocity.

46. A method according to claim 42, wherein obtaining the data signals comprises obtaining at least one signal from a microphone and a signal from a pressure sensor, and the blood pressure parameters comprise information on pulse wave shape.

47. A method according to claim 46, wherein obtaining at least one signal from the microphone comprises obtaining a low pass filtered signal and a high pass filtered signal from the microphone.

48. A blood circulation monitoring system adapted to be held around a part of a patient's body, the system comprising: a) an actuator or fluid-filled bag that exerts a controllable level of pressure locally on an artery in that part of the body, but does not exert pressure completely around that part of the body; b) a vibration sensor or microphone that presses against the artery to sense vibrations; and c) a processing unit that controls the level of pressure applied to the artery, and receives data from the vibration sensor or microphone.