WO2013110929A1

WO2013110929A1 - Aortic pulse wave velocity measurement

Info

Publication number: WO2013110929A1
Application number: PCT/GB2013/050131
Authority: WO
Inventors: Tobias Schaeffter; Phillipp BEERBAUM; Nicholas Richard GADDUM; Philip Jan Chowienczyk
Original assignee: King's College London
Priority date: 2012-01-26
Filing date: 2013-01-21
Publication date: 2013-08-01

Abstract

Embodiments of the present invention provide a measurement protocol that is able to non- invasively provide a degree of characterisation of aortic stiffness, at least qualitatively. The measurement protocol includes varying intra-thoracic pressure via a series of respiratory manoeuvres, whilst collecting aortic blood flow velocity information using a suitable scanning technique, and from which pulse wave velocity (PWV) may then be found. Plotting PWV against aortic trans-mural pressure variations caused by the changes in intra-thoracic pressure should allow a degree or aortic stiffness characterisation to be performed, in that a diseased aorta should show less variation in PWV with changes in trans-mural pressure than a healthy aorta. Also developed as another embodiment is a magnetic resonance (MR) technique for measuring aortic pulse wave velocity, which is able to achieve high temporal resolution on pulse velocity data capture, and hence lead to accurate PWV calculation. The technique uses a time sliding subtraction of velocity projections obtained from bipolar phase encoding gradients to produce velocity profiles at least every second repetition time T_R, and in some variants every T_R.

Description

Aortic Pulse Wave Velocity Measurement

Technical Field

The present invention relates to an aortic pulse wave velocity measurement technique, and to using pulse wave velocity to characterise aortic stiffness.

Background to the Invention and Prior Art

Hypertension contributes to approximately 50% of all cardiovascular deaths and is estimated to account for around 8 million deaths per annum worldwide [JJ. The vast majority of these deaths are attributable to systolic hypertension, this in turn being related to stiffening of the aorta and large vessels [2]. Arterial pulse wave velocity (PWV) has been shown to be indicative of arterial stiffness, and a good predictor of mortality in large cohorts [3]. However, the accuracy of currently available methods is not good enough for patient-specific [4] or blood-pressure-independent measures of aortic stiffness [5]. Thus aortic stiffness, a risk factor of paramount importance in this large patient population, cannot be reliably discriminated in the clinic.

PWV assessment is most accurately achieved using invasive catheter measurement, although the more common non-invasive means is indirect pressure pulse measurement at the carotid and femoral arteries using tonometers [6]. Although non-invasive measurements can be taken rapidly with high fidelity sensors to determine transit time between the two waveforms, uncertainty remains in the arterial length between these two sites [6]. Furthermore, in obese patients these methods are affected further by poor signal quality at the femoral site due to excessive adipose tissue.

Since early works by groups such as Boese et. al [7], groups have begun assessing PWV through phase contrast (PC) magnetic resonance (MR). The primary advantage using MR lies in the ability to accurately measure aortic length reducing geometry induced error in the PWV assessment. Retrospective gating of continuously acquired velocity information to the ECG over 2-3 minutes is averaged and interpolated to render a single cardiac cycle with sufficient temporal resolution and high signal quality [8]. Currently, PWV assessment using both in-plane, and through-plane 2D PC is performing well when compared to invasive pressure catheter data with associated errors of 0.1±0.8 m/s [9], and extending from global to regional assessments of stiffness [10]. Faster methods of PWV assessment using PC MR assessment methods project the velocity signal onto one dimension of the 2D velocity slice [11, 12] offering a reported temporal resolution of 14 ms [12]. To improve temporal resolution, these groups then stacked PC data over numerous staggered cardiac cycles [13],

Various studies have demonstrated the changeability of a subjects' PWV by affecting aortic trans-mural pressure through changes in arterial wall properties, arterial pressure, or intrathoracic pressure. These were achieved through respiratory manoeuvres [14], applied extracorporeal pressure [ 1 5], exercise [16], and pharmacological stimuli [5]. Furthermore, Stewart et. al demonstrated that although healthy individuals' PWVs varied through induced pharmacological vasoconstriction, hypertensives' PWVs remained unchanged under induced vasodilation [5].

Thus, accurate measurement of pulse wave velocity in vivo remains challenging. However, provided suitable measurement techniques can be found, with sufficient resolution, then pulse wave velocity may be used in various diagnostic or other characterisation techniques.

Summary of the Invention

Embodiments of the present invention provide a measurement protocol that is able to provide a degree of characterisation of aortic stiffness, at least qualitatively. The measurement protocol includes varying intra-thoracic pressure via a series of respiratory manoeuvres, whilst collecting aortic blood flow velocity information using a suitable scanning technique, and from which pulse wave velocity (PWV) may then be found. Plotting PWV against aortic trans-mural pressure variations caused by the changes in intra-thoracic pressure should allow a degree or aortic stiffness characterisation to be performed, in that a diseased aorta should show less variation in PWV with changes in trans-mural pressure than a healthy aorta. Also developed as another embodiment is a magnetic resonance (MR) technique for measuring aortic pulse wave velocity, which is able to achieve high temporal resolution on pulse velocity data capture, and hence along with accurate assessment of vessel length, can provide an accurate PWV assessment The technique uses a time sliding subtraction of aortic blood velocity projections obtained from bipolar phase encoding gradients to produce velocity profiles at least every second repetition time T_R, and in some variants every T_R. Various aspects of the invention are provided. In particular, from one aspect there is provided a method for characterising aortic stiffness, comprising: acquiring plural sets of aortic blood- flow velocity data, the sets being acquired under different aortic trans-mural pressures induced by respiratory manoeuvres, the sets comprising blood-flow velocity information from a plurality of aortic locations; calculating a pulse wave velocity from the sets of aortic blood flow velocity data; and plotting pulse wave velocity against aortic trans-mural pressure; wherein aortic stiffness may be characterised by the rate of change of pulse wave velocity with aortic trans-mural pressure.

With the above a method is provided that provides some information about aortic stiffness, in so far as variation of pulse wave velocity with trans-mural pressure may be found. In this respect, a healthy aorta is expected to exhibit marked change in PWV with intra-thoracic pressure changes, the intra-thoracic pressure changes being induced by the respiratory manoeuvres, and leading to changes in aortic trans-mural pressure, given a stable blood pressure. In contrast, a diseased aorta is expected to be stiffer, and exhibit less change in PWV for the same trans-mural pressure changes. As such a line plot with a shallower gradient, but a higher baseline may be obtained.

In one preferred embodiment the method further provides for plotting the pulse wave velocity data against aortic trans-mural pressure on a line graph to produce a graphical image. Hence, an image of the plot of PWV with trans-mural pressure, (as a surrogate for aortic wall strain with aortic wall stress), can be output, for subsequent interpretation by a clinician.

In one embodiment the aortic locations comprise two or more selected from the group comprising: the ascending aorta, aortic arch, supra-aortic branches and descending aorta until the diaphragm. In a preferred embodiment, three aortic locations along the ascending and descending aorta until the diaphragm are used.

The blood-flow velocity data may be acquired using one or more of: an ultra-sound scanner (Doppler), or a magnetic resonance imaging scanner. In a preferred embodiment an MR scanner is used, with a temporal resolution of around 11 ms, or more generally twice the repetition time. However, by using simultaneous dual slice excitation, a temporal resolution equal to the repetition time, (approximately 5.5 ms) can be obtained. In one embodiment the respiratory manoeuvres include one or more selected from the group comprising: free-breathing, normal breath hold, a Valsalva manoeuvre, and a Mueller manoeuvre. These manoeuvres have the advantage that they are easy to perform, induce acute changes in aortic wall strain without pharmacological intervention, and can be practised by a test subject in advance.

In one embodiment pulse wave velocity is calculated for a set of aortic blood flow velocity data by:

i) processing the set of aortic blood flow velocity data to obtain pulse velocity profiles of blood flow at the aortic locations;

ii) determining the transit time between two or more blood velocity waveforms;

iii) measuring distance between the aortic locations; and

iv) calculating pulse wave velocity in dependence on the determined transit times and the measured distance.

Determination of the transit times may take place in embodiments of the invention using any of the following:

a) finding the foot of each waveform and measuring time between found feet; or

b) cross correlating one of the following, and determining the transit time according to a correlation maximum:

a. complete waveforms;

b. data waveform up-slopes determined between the limits of a velocity data between a located velocity toward end diastole, and a located velocity toward the peak of systole; or

c. exponential or sigmoid profiles fitted to waveform up-slopes data determined between the limits of a velocity data between a located velocity toward end diastole, and a located velocity toward the peak of systole;

or

c) determining the transit time according to a minimum of absolute/squared errors between a portion, or all of the velocity waveforms. Although any of the above described means of measuring the transit time(s) of velocity waveforms may be used, the preferred embodiment is where the foot of each waveform is assessed and the transit times between waveforms is determined by measuring the time between these determined transit times. In this embodiment the pulse foot locations are found as the intersection of a tangent of the point of maximum gradient on the pulse upslope, and a tangent of the point of diastolic minimum.

From a second aspect there is provided a method for determining pulse wave velocity in a human aorta, comprising:

performing magnetic resonance scanning upon one or more thoracic transverse slices positioned so as to intersect at least two aortic locations;

subtracting phase encoded projections from successive scans of the same slice to obtain one-dimensional velocity profiles;

determining a pulse velocity profile per aortic location;

determining transit times between pulses in the pulse velocity profiles; and

calculating pulse wave velocity in dependence on the transit times and aortic length between the aortic locations.

In one embodiment the subtracting further comprises subtracting a phase encoded velocity projection obtained from a most recent scan of a slice from a phase encoded velocity projection obtained from a scan of the same slice immediately preceding in time the most recent scan, wherein the subtraction is repeated from scan to scan such that a one - dimensional velocity profile is obtained per subsequent scan.

In one embodiment the one dimensional velocity profiles are conveniently stored as lines in an M-mode image; and the M-mode images are processed to determine the pulse velocity profiles.

In embodiments of the invention the scanning comprises any one selected from the group comprising:

i) scanning using bipolar phase encoding gradients, a sign of which is toggled between successive scans of a slice;

ii) scanning using a zero then non-zero phase encoding gradient which are then toggled between successive scans of a slice; or iii) scanning using either a reference and then successive bipolar or zero/non-zero phase encoding gradients which are subtracted from the reference image.

In one embodiment preferably two thoracic transverse slices are scanned, an M-mode image being produced per slice, a one dimensional velocity profile being stored in the M mode image relating to the slice in respect of which the velocity profile was found.

In addition, in one preferred embodiment two thoracic transverse slices are scanned, the scanning of the two slices being time interleaved, wherein a first slice is scanned using two phase encoding modes, followed by the second slice being scanned using two phase encoding modes.

In another particularly preferred embodiment two thoracic transverse slices are scanned, the scanning of each slice being performed in parallel, and determination and storage of one - dimensional velocity profiles also being performed in parallel. This has the distinct advantage that temporal resolution is improved.

In embodiments of the invention the determination of velocity profiles is performed substantially in real-time with the scans.

In embodiments of the invention determination of the transit times uses any of the following techniques:

a) finding the foot of each waveform and measuring time between found feet; or

a. complete waveforms;

c. exponential or sigmoid profiles fitted to waveform up-slopes data determined between the limits of a velocity data between a located velocity toward end diastole, and a located velocity toward the peak of systole; or

c) determining the transit time according to a minimum of absolute/squared errors between a portion, or all of the velocity waveforms.

From a further aspect there is provided a method of characterising aortic stiffness according to the above described first aspect, wherein pulse wave velocity is found using a method according to the above described second aspect.

Finally, a further aspect provides an MRI scanner arranged to operate to find aortic pulse wave velocity according to the method of the second aspect above.

Brief Description of the Drawings

Figure 1 is a visual description of the interleaved, real time RTP scan M-mode image generation of an embodiment of the invention: Two slices are sequentially excited and a subtractions of the velocity projections provide one-dimensional encoding of velocity every 2T_R for each slice. These projected velocity vectors are then stacked together providing M- mode projections for each slice (1 and 2).

Figure 2 is a photograph of an MRI compatible mock ventricle and aorta pulsatile phantom, where; 1, reciprocating ventricle piston; 2, ventricle and working fluid; 3, arterial valve; 4, suspension fluid; 5, outer cylindrical case; 6, venous reservoir; 7, silicone aorta; 8, venous channel; 9, venous resistance screw; 10, arterial resistance screw and arterial catheter access; 11, arterial compliance chamber; 12, venous valve.

Figure 3 shows graphs illustrating comparison of pressure catheter derived PWVs to 2D PC and RTP MR.

Figure 4(A) is an oblique sagittal survey slice with RTP slice selections overlay ed

Figure 4(B) illustrates selected ascending and superior descending velocity profiles imposed above the flow encoded M-mode projected image Figure 5 shows human aortic velocity profiles at the ascending, superior descending and inferior descending aorta with located maximum gradients, minimum radii of curvature and wave 'feet' of the 2D PC (Fig 5 A) and real time RTP (Fig 5B) MRI scans.

Figure 6 shows correlation (Fig 6A) and Bland Altman (Fig 6B) comparisons between PWV derived from 2D PC and RTP MRI, the latter annotated with the mean PWV difference, and 1.96 standard deviations. Fig 6C is a graph of inter-scan variability.

Figure 7 shows beat-to-beat PWV (dotted line) during normal breath hold (top, Fig 7A), Valsalva manoeuvre (middle, Fig7A) and Miieller manoeuvre (bottom, Fig 7A), together with projected velocity profiles at the ascending and lower descending aorta (solid lines), as well as mean PWV ± one standard deviation for the three volunteers (Fig7B).

Figure 8 is a diagram illustrating aorta wall constructions;

Figure 9 A shows a plot of stress strain response for an artery wall;

Figure 9B illustrates a plot of pulse wave velocity against trans-mural pressure from a healthy aorta;

Figure 9C shows a plot of pulse wave velocity versus trans-mural pressure for a stiff or unhealthy aorta;

Figure 10 is a flow diagram of method steps according to a first embodiment of the invention; Figure 11 is a block diagram of apparatus elements of an embodiment of the invention;

Figure 12 is a flow diagram of process steps involved in an embodiment of the present invention;

Figure 13 is a flow diagram of method steps involved in an embodiment of the invention; and Figure 14 is a graph illustrating calculation of "feet" of a velocity profile used in an embodiment of the invention.

Description of the Embodiments

The present invention has a number of embodiments. In a first embodiment, to be described in more detail later, a measurement protocol is presented which aims to provide for characterisation of aortic stiffness, by measuring changes in pulse wave velocity with changes in trans-mural pressure. Specifically, a healthy aorta is thought to exhibit significant changes in pulse wave velocity with trans-mural pressure, whereas an unhealthy or stiff aorta is thought to exhibit much less significant change in pulse wave velocity, with changes in trans-mural pressure. In the measurement protocol to be described, changes in trans-mural pressure are obtained by performing various respiratory manoeuvrers whilst measurement of aortic velocity flow is taken. In particular, trans-mural pressure may be decreased by performing a Valsalva manoeuvre i.e. forced expiration against a closed glottis, whereas trans-mural pressure may be increased by performing a Mueller manoeuvre i.e. forced inspiration against a closed glottis. When used in the process of one embodiment of the invention, these manoeuvres will be typically applied against a mouth piece which is connected to a pressure transducer via a sealed conduit. This will provide the intra-thoracic pressure of the subject.

A second embodiment of the invention describes a pulse wave velocity measurement technique, using phase contrast magnetic resonance imaging. The pulse wave velocity measurement technique is particularly suitable for use in the measurement protocol of the first embodiment, although may also find application elsewhere, where real time aortic pulse wave velocity is required to be known.

A third embodiment, based upon application of the pulse wave velocity measurement of the second embodiment together with the measurement protocol of the first embodiment to actual volunteers will also be described.

Discussing in detail the first embodiment relating to the measurement protocol, Figure 8 shows a diagram of an example artery. The artery comprises an artery wall 80, and blood pressure 82 within the artery acts on the artery wall in an outwards direction. This is countered by intra-thoracic pressure 84 acting against the artery wall from the outside. The aorta is the primary artery leading from the heart. A healthy aorta wall primarily constitutes flexible elastin and stiffer collagen. In contrast, a diseased aorta typically has significant lipid or cholesterol or plaque build up, along with inflammation of the wall tissue itself. The build up along the internal surface of a diseased aorta is thought to affect the material properties thereof, and to make the wall "stiffer". However, characterisation of aortic stiffness has heretofore been difficult in vivo.

The present embodiment relies on measurements of pulse wave velocity along the aorta whilst varying trans-mural pressure across the aorta wall, and therefore aortic wall hoop strain. In this respect, intra-thoracic pressure is varied using respiratory techniques. For example having the subject perform a Valsalva manoeuvre will increase intra-thoracic pressure, and hence decrease trans-mural pressure. This should have the effect of reducing pulse wave velocity in the aorta, provided the aorta is able to respond to the reduced transmural pressure by not being hampered by deposits along the interior wall. Transversely, increasing trans-mural pressure i.e. by reducing intra-thoracic pressure using a Miieller manoeuvre or the like, is expected to have the effect of increasing pulse wave velocity. Again, however, the degree of pulse wave velocity increase for a given increase in transmural pressure will also depend upon the overall stiffness of the arterial wall, depending in turn upon plaque deposits, aortic wall inflammation and the like thereon. The collected PWV data will then be plotted against their aortic trans-mural pressure. This data will then be fit to a suitable characterisation function, (for example, but not limited to, an exponential/polynomial function). The characterising constants of such a function may be used as diagnostic parameters in a subsequent clinical diagnosis.

Figures 9B and C illustrate a possible discriminating pair of changes in pulse wave velocity with trans-mural pressure plots for a healthy aorta, and an unhealthy "stiff aorta respectively. As will be seen from Figure 9(B) pulse wave velocity varies significantly with trans-mural pressure in a healthy aorta, whereas in an unhealthy "stiff aorta, there is little change in pulse wave velocity with trans-mural pressure. Therefore, by performing a combination of respiratory manoeuvres, such as Valsalva and Miieller manoeuvres to alter intra-thoracic pressure, trans-mural pressure can be varied over a remarkable range. Consequently, the effected pulse wave velocity can wander up and down a curve which characterises the relationship between trans-mural pressure (or arterial hoop strain) and pulse wave velocity. By measuring pulse wave velocity whilst performing such respiratory manoeuvres to vary trans-mural pressure across the aortic wall, it becomes possible to obtain a characterisation of the aorta in dependence on pulse wave velocity and trans-mural pressure, to determine whether an aorta is healthy, or subject to significant deposits which stiffen the aorta wall.

Figure 10 illustrates a measurement protocol to be followed to allow such aortic characterisation. The protocol is divided into two parts, a first data acquisition phase where the test subject or patient is present, and a second post processing phase, where various calculations must then be performed on the acquired data, and output images produced. From the output images a determination of aortic stiffness may then be obtained. This determination may then be subsequently applied together with other information if needed by a clinician to determine a diagnosis of a particular disease. As such, therefore, the present measurement protocol does not lead per se to specific diagnosis of any particular disease, but instead provides information from which aortic stiffness may be characterised, this characterisation then being usable subsequently for diagnostic purposes.

Referring to Figure 10, at step 10.2 the test subject and equipment is prepared. In this respect, the purpose of the measuring equipment is to allow aortic blood-flow velocity to be found, where the velocity is calculated from velocity data taken from various different parts of the aorta, together with a knowledge of aortic length between the parts at which the velocity was sampled. A set of aortic velocity data comprising velocity measurements at different parts of the aorta must therefore be acquired under different intra-thoracic pressure conditions. Pulse wave velocity can then be calculated for each velocity data set, and then plotted against trans-mural pressure to provide a graph which characterises aortic stiffness.

In view of the above, any suitable imaging or measurement equipment that is able to measure the velocity of blood flow in the aorta may be used. This includes magnetic resonance imaging scanners, as will be described in more detail later with respect to other embodiments, but may also include other scanning technologies, such as a Doppler ultrasound scanner. All that is important is that the scanning equipment used is able to obtain a measurement of blood flow velocity within the thoracic aorta at two or more different locations. These need to be measured concurrently, or separately and gated to the ECG (electrocardiogram).

Assuming after step 10.2 the test subject and the measurement equipment is ready, the test subject is then instructed to breathe freely or maintain a normal breath hold, and at step 10.4 a first set of aortic velocity data is acquired using the measurement equipment. For example, the measurement equipment is used to obtain velocity data for several aortic locations, such as the ascending, upper descending, and lower descending aorta. Multiple velocity measurements may be taken as required, or only a single set. The velocity data acquired at step 10.4 constitutes the free breathing/normal breath hold velocity data set, from which pulse wave velocity can then be calculated corresponding to, for example, point A in Figures 9B and C. After the first set of velocity data has been acquired, the test subject is then instructed to perform respiratory manoeuvres such as a Mueller or Valsalva manoeuvre, or a combination of the two. This has the effect described previously of changing trans-mural pressure, and whilst the patient performs the manoeuvre a second set of aortic velocity data is acquired. Again, the second set of data comprises velocity measurements at the ascending, upper descending, and lower descending aorta. One or more such sets of data may be acquired.

After the second set of aortic velocity data has been acquired with a higher trans-mural pressure, the test subject is then instructed to perform a further respiratory, such as a Valsalva manoeuvre, or the like, in order to lower trans-mural pressure. Whilst the Valsalva manoeuvre is being performed, at step 10.8 a third set of aortic velocity data is acquired, again from the same aorta locations. One or more such sets of data may be required.

It will be appreciated that the above respiratory manoeuvres are described by way of example only, and may be performed in a different order than that described, and in fact any order or combination of the respiratory manoeuvres may be performed. For example, the free breathing measurements may be taken last and likewise the measurements with the increased or decreased trans-mural pressure may be performed in any order. In practice it will likely be necessary to acquire velocity profiles over a range of trans-mural pressures. This will likely require one or repeated scans where the subject is asked to perform (for example) a combination of respiratory manoeuvres. The desired outcome of the respiratory protocol used is to sample blood velocity over a large range of aortic trans-mural pressures.

In addition, whilst the aortic velocity data is being obtained measurements of intra-thoracic pressure are also taken at the same time, via the mouth piece connected to a pressure transducer. Where possible, blood pressure is also recorded, to allow an accurate trans-mural pressure to be calculated. If blood pressure cannot be conveniently recorded (for example because the test subject is in an MR scanner), then the respiratory techniques should be chosen to be performed for a time or in combination that effect on blood pressure is negligible, or else is known in advance.

Once sufficient sets of velocity data vs. aortic trans-mural pressure data has been obtained to facilitate the characterisation of the patient's aorta elastic mechanics using a pre-defined function, the test subject may then be released, at step 10.10. The test subject is then free to leave, and a data acquisition phase which requires the presence of the subject is finished.

A post processing phase then commences, in order to process the acquired data in order to firstly calculate pulse wave velocity for each velocity data set, and then secondly to plot pulse wave velocity against trans-mural pressure within a graphical image of the plot. The graphical image can then be used to characterise aortic stiffness, for example by fitting the data to a predefined function (for example, but not limited to, an exponential/polynomial function. The characterising constants of such a function may then be used as diagnostic parameters in a subsequent clinical diagnosis, but it should be noted do not themselves provide diagnosis of any specific disease..

The first steps that are performed in the post processing phase are to calculate the pulse wave velocity for each velocity data set. A processing loop is started at step 10.12, wherein the pulse wave velocity for a data set is calculated at step 10.14. An evaluation is performed at step 10.16 as to whether each velocity data set has been calculated, and if not the pulse wave velocity is calculated for that data set.

Calculation of pulse wave velocity is performed preferably using the method shown in Figure 13, from steps 13.6 onward. This assumes that each velocity data set comprises a plot of data mapping velocity at a particular aortic position against time. When a pulse wave travels along the aorta, the velocity at a particular aortic position will increase and then subsequently decrease as the pulse passes. Figure 14 illustrates a plot of velocity against time of a pulse wave front. A first pulse wave front 140 is measured at a first aortic location, and a second pulse wave front 142 is measured at a second, subsequent, aortic location in the direction of velocity flow. To calculate pulse wave velocity, in this embodiment a pulse wave form is analysed to determine the point of maximum upward gradient, shown as 1442 for pulse waveform 140. In addition, the point of local mean of the diastolic minimum is also found, shown in Figure 14 as .1462 for pulse waveform 140. Gradient line 144 is extrapolated from the point of maximum gradient 1442, and this intersects a line extending from the local mean of the diastolic minimum shown on Figure 14 as 146. The intersection of lines 144 and 146, shown as .148, constitutes what is referred to as the "foot" of the pulse 140, and this foot then forms the basis of the calculation of pulse wave velocity for the pulse. In addition, the same calculations are performed for the pulse 142, taken at the subsequent aortic location i.e. the point of maximum upward gradient is found, as well as the mean of the diastolic minimum. The intersection of lines extending from these two points then constitute the foot of pulse 142, shown as point 150. Finding the "feet" of pulses 140 and 142 therefore allows a time difference between points 150 and 148 to be found. By then measuring the length of aorta between the two measurement points or "feet" relating to the pulses 140 and 142, for example on an image of the aorta acquired by ultrasound or MRI scanning, the velocity of the pulse between the two measurement positions can be found. This velocity then constitutes the pulse wave velocity for the data set.

It should be noted that in other embodiments other techniques for determining the pulse wave transit times may be used, other than the "feet" technique described above. Various techniques to determine pulse wave velocity from pulse velocity waveforms are known in the art.

Once the pulse wave velocity has been found for each data set, at step 10.18 it is plotted against trans-mural pressure, to obtain a graphical plot of change in pulse weight velocity with trans-mural pressure. In this respect, trans-mural pressure can be found by monitoring blood pressure of the user while the data is acquired, and measuring intra-thoracic pressure by use of a mouthpiece connected to a pressure transducer, and recording the data obtained If the blood velocity assessments are carried out in an environment (such as a MR scanner) which reduces access to blood pressure measurement, the respiratory manoeuvres can be carried out such that either i) the blood pressure has minimal change, or ii) the blood pressure changes are predictable. By then fitting a predefined function (for example, but not limited to, an exponential/polynomial function), to the pulse wave velocity vs. transmural pressure data, the subject's aortic mechanics may be characterised, at step 10.20. The constants of such a function may be used subsequently as diagnostic parameters. In this respect, a diseased aorta may exhibit a shallow gradient indicating little change in pulse weight velocity with changes of trans-mural pressure. An example image plot is that shown in Figures 9B, or 9C.

With the first embodiment, therefore, a measurement protocol is provided which allows for aortic characterisation to allow distinguishing between a healthy aorta and a diseased aorta m vivo, and without any invasive procedure. As noted above, the technique depends upon accurate measurement of pulse wave velocity, but is not dependent upon any particular pulse wave velocity measurement technique. Moreover, although we have described calculation of pulse wave velocity from arterial velocity data using the technique of finding the "foot" of each pulse, other pulse wave velocity calculation techniques are known in the art, which may also be used. Important aspects of the first embodiment are therefore the variation of transmural pressure using respiratory techniques, and the calculation of pulse wave velocity for each different trans-mural pressure thus obtained. Plotting the pulse wave velocities thus obtained against trans-mural pressure gives a characterisation of the aorta, which can be used to determine the health of the aorta. Such information may then be used by a clinician is subsequent diagnoses of particular diseases.

A second embodiment will now be described, which focuses on accurate calculation of pulse wave velocity. The pulse wave velocity found can be used, for example in the measurement protocol of the first embodiment, or may be used for other purposes.

The second embodiment provides a magnetic resonance (MR) scan technique which is able to obtain blood flow velocity measurements at a high temporal resolution, and thus provides good velocity data from which pulse wave velocity can be calculated.

Figure 1 1 shows a typical MRI apparatus, comprising MRI scanner 110, into which is placed subject 120 to be scanned. A control computer 114 controls MRI scanner 110, and is provided with an MRI interface 1122, to control the scanner, and receive data therefrom. A central processing unit 1 124 is provided, as well as memory 1126. The computer is provided with a display interface 1128 connected to a display screen, and an input interface 1130 connected to a keyboard or the like, for receiving control commands. Long term data storage 116 such as a hard disk or the like is provided, on which data such as MRI image data 1162 may be stored. In addition, control programs to be executed by the CPU 1124 to control the MRI scanner and to process the obtained data are also provided. In particular, MRI scan acquisition program 1164 controls the MRI scanner 110 via the MRI interface 1122 to perform an actual scan and obtain data, whereas MRI image processing program 1166, when executed by the CPU 1124, can process the obtained MRI image data 1162 in order to determine characteristics thereof, in particular in this embodiment to calculate pulse wave velocity. Figure 1 shows further details of the MRI scan to calculate pulse wave velocity. Firstly, two transversal slice selections 1 and 2 are made, the first slice being positioned at the ascending aorta, and the second slice being positioned just about at the LV apex. The first slice therefore obtains velocity profiles from the ascending aorta and the upper descending aorta, whereas the second slice is able to obtain velocity data from the lower descending aorta.

Figure 12 shows a flow diagram of the scanning order of slices, and processing that is performed to obtain velocity projection data. In this regard, Figure 12 represents the various slice subtractions shown in Figure 1, and described further below.

In steps 12.2 and 12.4, slice 1 is scanned twice using bipolar phase encoding gradients employed in a standard gradient-echo sequence for velocity encoding. In this sequence only the anterior posterior (AP) field of view (FOV) was encoded with the spin frequency and fold over direction orientated in the left to right (LR) direction so that only one dimensional (ID) velocity projections were obtained after each excitation. The sign of the bipolar gradients were toggled between excitations (i.e. between steps 12.2 and 12.4) and then at step 12.6 subtractions of the subsequent projections results in a one-dimensional velocity profile, along the anterior to posterior direction. The velocity profile is then stacked in an M mode image for slice 1, see Figure 1 for an example of such an image. Having scanned slice 1 twice using the different bipolar gradients, slice 2 is then scanned in the same manner, again with a first bipolar gradient line, and then again with the opposite bipolar gradient sign, at steps 12.10 and steps 12.12. The resulting velocity projections are then subtracted one from the other at step 12.14, to determine a velocity profile for slice 2. This is stored in a respective M mode image for slice 2 at step 12.16 (again, see Figure 1). After calculation of the first velocity profile for the first slice and the second slice, the two slices are then scanned again in order in turn, again with toggled bipolar phase encoding gradients. Therefore, as additional slice velocity data is obtained, additional subtractions can be performed to obtain additional velocity profiles, as shown in Figure 1.

For example, as shown in Figure 12, at step 12.18, slice 1 is scanned with the first bipolar gradient sign, and then at step 12.20 a second velocity subtraction subtracting the results of the scan performed at step 12.4 from that performed at step 12.18 can be obtained to determine a second velocity profile for slice 1. This second velocity profile is then temporally stacked in the slice 1 M mode image at step 12.22. Slice 1 is then scanned again with the second bipolar gradient sign, and a further subtraction of the velocity projection obtained at step 12.24 from the velocity projection obtained at step 12.18 is performed to determine a third velocity profile for slice 1, at step 12.26, and this is then temporally stacked in the M mode image for slice 1, at step 12.28. Slice 2 would then be scanned again, and similar processing steps to steps 12.20 to 12.28 performed for slice 2, with respect to the slice 2 velocity projections previously obtained at steps 12.10 and 12.12. This way a time-sliding subtraction is obtained to give ID projected velocity profiles that are then stacked in time to form the M mode images for each slice. Example M mode images are shown in Figure 1. Given that a pair of ID bipolar read outs is generated for both slices every four repetition times (4T_R) and that subtractions are made within bipolar gradient pairs, a velocity encoded ID profile is obtained for each slice every second repetition time. For a repetition time of 5.5ms, this means that a temporal resolution of 1 1ms is obtained to sample the velocity of aortic blood flow at each aortic position.

Within the above embodiment bipolar phase encoding is used to obtain velocity information. In other embodiments, however, other methods of phase encoding may be used to obtain velocity sensitivity. Various such phase contrast techniques to obtain blood velocity information using various MR sequences are known in the art.

Having acquired M mode images for slices 1 and 2, and which encode the velocity data for each aortic position the slices pass through, it then becomes necessary to process the M mode images in order to obtain calculations of pulse wave velocity. Processing of the M mode images can be performed after image acquisition has been obtained, and hence does not require the presence of the test subject. Processing of the images in the present embodiment is performed by the MRI image processing program 1166, further details of which are shown in Figure 13.

With respect to Figure 13, at step 13.2 the image processing program 1166 accesses the M mode images for slices 1 and 2 generated by the image acquisition program. It is then necessary to process these images to extract velocity information therefrom. In this respect, the M mode image for slice 1 contains two sets of velocity data, for the ascending aorta, and the upper descending aorta. This velocity data is shown in the M mode image for slice 1 in Figure 1. The M mode image for slice 2 contains the velocity data for the lower descending aorta. In order to calculate pulse wave velocities, it is necessary to extract the velocity profile data from the M mode images to obtain pulse velocity profiles such as those shown as pulses 140 and 142 in Figure 14. In order to achieve this, at steps 13.4 a one dimensional spatial kernel is applied along the read out direction (anterior to posterior) of the M mode images. For the M mode image from slice 1, which contains two sets of velocity data, the one dimensional spatial kernel is applied to those parts of the image representing the ascending aorta and the upper descending aorta separately, respectively. Applying such one dimensional spatial kernels to these images then obtains velocity profile data from the image, being a plot of velocity against time for each of the three aortic positions. For example, Figure 5B illustrates three aortic velocity profiles for the ascending aorta, the upper descending aorta, and the lower descending aorta.

Having obtained the velocity profile for each position, at step 13.6 a processing loop is started to process each velocity profile to determine the "foot" of the velocity profile. Steps 13.8 to steps 13.12 are then performed on each velocity profile to determine the foot, in the same manner as described previously with respect to Figure 14 in the first embodiment. That is, each respective foot point 148 or 150 is defined as the intersection of the maximum gradient of the upslope according to a moving average window (where n usually equals 4), and a horizontal projection from the local mean (n=5) about the end diastolic minimum. Figure 14 illustrates this, as described previously, and a further example is also shown in Figure 5B.

Given the located "feet" for each pulse velocity profile, time differences between the velocity profile "feet" are measured at step 13.16. In Figure 14, the time difference represents the period of time between each successive velocity profile "foot". Knowing the time difference between each pulse at each aortic position, it then becomes necessary to measure the aorta length, and this is performed at step 13.18, for example from MRI imagery of the aorta. An MRI geometry scan to allow for aorta length assessment is shown in Figure 4A. Once the aortic length is known, it then becomes possible to find the pulse wave velocity between the different aortic segments, by dividing the measured distance between the aortic positions at which blood flow velocity measurements were taken by the time difference between the respective "feet" of the velocity profiles for each respective aortic position. As noted, with a repetition time of 5.5ms, the temporal resolution of the velocity profiles becomes 11ms, and hence a very accurate pulse wave velocity can be found. Beat to beat pulse wave velocity may be found, as well as mean pulse wave velocity over a time period, such as ten cardiac cycles. The advantages of this technique are described in further detail later with respect to the third embodiment, but such a scan technique to obtain pulse wave velocity shows good repeatability, with an inter-scan standard deviation of 0.43 plus or minus

0.29 metres per second.

Modifications may be made to the above scan routine to provide improvements, particularly in temporal resolution. For example, in the above described embodiment whilst dual slices are used, and slices excited twice in turn, with toggled bipolar gradients for each excitation. However, in an improved technique, simultaneous dual slice excitation is used, i.e. both slices 1 and 2 are excited at the same time, and temporal resolution can be improved by a factor of 2

1. e. a temporal resolution for the velocity profile of 5.5 ms should be achievable. In this respect, by using a simultaneous dual slice excitation, rather than the slices being temporally interleaved as in Figure 1, the slices are essentially being temporally processed in parallel, and hence a 1-D velocity projection by virtue of the sliding subtraction of the most recent velocity projection from the previous velocity projection becomes available every repetition time T_R. Improved temporal resolution in the pulse velocity profile ultimately leads to increased accuracy in pulse wave velocity calculation.

In a further modification, given that in order to find pulse wave velocity it is necessary to have a pulse velocity profile from at least two locations in the aorta, and to know the distance between those locations, in a variation of the second embodiment instead of using two transverse slices, only slice 1 is excited, and pulse wave velocity is calculated between the ascending and upper descending aorta only. In such a modification, the processing performed is identical to that described previously, except without performing slice 2 excitation and signal processing. Therefore, there is no need to time interleave slice 2 excitations into slice 1 excitations, and hence temporal resolution for ID velocity projections for slice 1 can be increased such that a velocity projection is found every repetition time.

Finally, as in the first embodiment, it is not essential to use the technique of finding pulse "feet" to determine pulse transit times, and other techniques for determining pulse transit times are known, which may be used in alternative embodiments. Various other techniques were mentioned previously in the summary portion.

A third embodiment of the invention will now be described. The third embodiment combines the pulse wave velocity calculation using novel MRI scan of the second embodiment, with the measurement protocol of the first embodiment, and also presents results obtained from live subjects. The third embodiment will be described with respect to Figures 1 to 7, discussed briefly previously. The third embodiment is described further below.

As with the first embodiment, the third embodiment aims to realise individual characterisation of aortic stiffness by developing a means of accurately assessing beat-to-beat PWV. For this embodiment we developed a new, MR based, real-time projected velocity (RTP) scan which images real time blood velocity in the thoracic aorta. The aim of this study was to assess the accuracy of RTP-derived PWV, and to investigate whether beat-to-beat PWV variations through changes in intra-thoracic pressure can be observed. We employed a pulsatile flow phantom to compare pressure catheter-derived PWV assessment with those measured with both the new RTP scan and the current MR standard 2D PC at 3 T MRI. The RTP scan's performance in vivo was then assessed in a volunteer study, including an evaluation of the variability of PWV through changes in intra-thoracic pressure during Valsalva and Mueller manoeuvres.

Methods

The Real Time Projection (RTP) Phase-Contrast MR-Scan

Bipolar phase-encoding gradients were employed in a standard gradient echo sequence for velocity encoding. In this sequence no phase-encoding was applied so that velocity projections, (projected in the right to left direction), were obtained after each excitation. The sign of the bipolar gradients were toggled between excitations and subtraction of the subsequent projections resulted in a one-dimensional velocity profile, (along the anterior-to- posterior direction). Over successive read outs, an M-mode type image was built up with anterior-to-posterior field of view along the vertical axis, and time along the horizontal axis. M-mode velocity projections were obtained at two slice locations in order to assess a pulse wave transit time. Figure 1 presents a schematic of the RTP scan; from slice selection (left), sliding subtraction and example m-mode images (right). Transversal slice selections (1 and 2) are positioned at the ascending aorta and just about the LV apex. Pairs of bipolar gradient readouts are interleaved in real-time. Then, 1-D projected velocity profiles are generated from sliding subtractions; i.e. for each slice, the first pair of bi-polar gradient readouts are subtracted, then the last and first lobes of the first and second bi-polar gradients respectively are subtracted, etc. Each of these 1-D projected velocity profiles are stacked in time to form the M-mode images, examples are shown in the figure. Example M-mode projected velocity images at both slice locations after sliding subtractions have been performed. Considering that a pair of 1-D bipolar readouts is generated for both slices every four repetition times, (4T_R), and that subtractions are made within bipolar gradient pairs, a velocity-encoded 1-D profile is obtained for each slice every 2T_R.

In order to assess the efficacy of PWV measurements using the RTP scan, a pulsatile flow phantom study was carried out where these measurements were compared to those obtained using standard 2D phase contrast (PC) MRI, as well as PWV derived from the gold standard, pressure catheter measurements. Following this, a volunteer study was conducted where the two MRI protocols were compared. Finally, volunteers were asked to perform Valsalva and Miieller manoeuvres during the RTP scan to observe its ability to observe beat-to-beat PWV variations during acute changes in aortic trans-mural pressure.

Pulsatile Flow Phantom

The phantom's ventricle is driven by a linear servo actuator (ETB-32, Parker Hannifin) and piston assembly, located outside the 5 Gauss line coupled to the ventricle piston via a 2 m rigid boom. The piston ejects water through a tri-leaflet polyurethane arterial valve, (Hemolab, Eindhoven, The Netherlands), and into a 450 mm, 018.5mm silicone aortic tube, (TU Eindhoven, The Netherlands), see Figure 2. Afterload is maintained by a 3-element Windkessel at the distal end of the aortic tube, after which the working fluid returns via a venous reservoir and valves back to the ventricle. Closed-loop feedback control of the ventricular action was maintained via a multi-function I/O board and a servo controller using Lab VIEW software, (PCI-MI016-E4 and PC-servo-4A, National Instruments, Austin, TX, USA).

Pressure measurements at two locations were taken outside the scanner using two 6F fluid filled pressure catheter, (Arrow, Reading, PA, USA), and pressure transducer (Omega, Irlam, Manchester, UK), combinations. The analogue signal from the transducers were acquired at 100 Hz via a data acquisition card, (USB-6009, National Instruments, Austin, TX, USA), and processed and recorded using in-house software, (The MathWorks, Natick, MA, USA). The catheters were left in place for the MR-acquisition. The rig was run at 60 BPM, and the test repeated four times, each time adjusting the arterial resistance screw to change the afterload.

MR Acquisition

Twenty MR data sets comprising both 2D PC and RTP MRI scans were obtained from fifteen healthy volunteers (Table 1). On each volunteer the RTP scan was repeated three times to observe intra-scan PWV variability. Additional scans during respiratory manoeuvres were performed on six volunteers to assess the RTP scan's ability to observe PWV variation. Upon each of these volunteers, three repeated scans were carried out while performing Valsalva manoeuvres (positive intra-thoracic pressure through expiration effort against a closed glottis), and another three performing Mueller manoeuvres (negative intra-thoracic pressure through inspiration effort against a closed glottis).

Mean ± Std

Age (yrs) 32 ± 7

Height (cm) 181 ± 9

Weight (kg) 77 ± 14

Brachial Diastolic Pressure (mmHg) 72 ± 7

Brachial Systolic Pressure (mmHg) 120 ± 7

Brachial Mean Pressure (mmHg) 88 ± 6

N (males) 15 (10)

Table 1

MR-acquisitions were performed on a 3T scanner, (Philips Achieva, Philips Healthcare, Best, The Netherlands). A 32-element RF coil was used for signal reception. Initial surveys were used to plan the scan orientation to ensure accurate through-plane velocity encoding. The upper and lower slices of the RTP scan were placed above or below the level of the pulmonary artery and just above the left ventricle (LV) apex respectively (Figure 1 left). The RTP scan is based on an un-gated T1TPC technique to acquire velocity encoded projections during a 10 second long breath-hold. Velocity encoding was set along the feet-head direction, using velocity encoding between 120 and 150 cm/s and following acquisition parameters: T_E/T_R = 3.1/5.5 ms, flip angle of 12°, voxel size of 1.8 [AP] ^χ 8.0 [FH] ^χ 320 [RL] mm³ (projection along the RL direction). The projections of the two slices were spaced at 11ms intervals (2T_R) which represents the temporal resolution of the PWV measurements. For comparison, a free breathing, in-plane, 2D phase contrast scan was performed [9]. Velocity encoding was performed in the feet-head direction, with a velocity sensitivity of 150 cm/s, a voxel size of 1.8 x 2.5 x 8.0 mm³, 2 signal averages (NSA), T_E/T_R = 2.9/5.3 ms and a flip angle of 12°. The scan was retrospectively gated to acquire 100 phases within one averaged cardiac cycle over two and three minutes (depending on heart rate).

As two NSA were used with in-plane 2D PC imaging, the mean pulse wave velocities of three RTP scans of 10 seconds each were used for comparison, and their standard deviation recorded to indicate inter-scan variation.

Systolic and diastolic pressure was recorded by brachial oscillometry (Datex Ohmeda Patient MRI Monitor Oscillometer, GE Healthcare, Buckinghamshire, UK) at the end of each volunteer's scan. This was carried out while the volunteer remained in the supine position on the scanner table.

Post processing

The aortic length between slice selections was measured from oblique sagittal slices encompassing the thoracic aorta (Figure 4, left).

Pulse wave velocities obtained in real-time by the RTP scan and by the conventional in-plane 2D phase-contrast scan were processed by use of a customized program written for each method (MATLAB, The MathWorks, Natick, MA, USA). Within either, regions-of-interest masks were manually selected to identify transient velocity profiles. The masks were applied at the two slice locations along the silicone tube in the case of the phantom, and at the ascending, upper descending and lower descending aorta in the volunteers. These masks were either ID spatial kernels along the read-out direction (AP) of the RTP images, (see Figure 1), or circular, prescribing the lumina in the case of 2D PC images. The velocity data within the lumen was then averaged for each time step providing transient blood velocity profiles. Transit time calculations for each of the velocity profiles was achieved by way of the foot-to-foot transit-time method [6]. The foot was defined as the intersection of the maximum gradient of the upslope according to a moving average window (n = 4), and a horizontal projection from the local mean (n = 5) about the end diastolic minimum.

Given the located 'feet', transit times between the three aortic sections for each of the cardiac cycles were then measured, (i.e. 10 cardiac cycles at 60 BPM). Occasionally false 'feet' would be located by the processor due to either projected velocity signal from the pulmonary artery in the upper slice, or the liver in the lower slice. If the transit time fell outside an acceptance envelope with reference to the median transit time, (t_med), of 0.5t_med to 2t_med then it was excluded from further calculations. Considering the measured vessel length, it was then possible to calculate both the beat-to-beat PWV and the mean PWV over the ten second scan. In-plane 2D PC and RTP scans were carried out on each volunteer and pulsatile phantom scan.

Statistics

Statistical analysis was carried out in Matlab using the functionality from its Statistical Toolbox. Curve fitting was also carried out in Matlab using a Least Squares approximation. All continuous data were expressed as mean ± 1.0 standard deviation. Bland Altman comparisons are made as mean ± 1.96 standard deviations as per convention.

Results

Validation Using the Pulsatile Phantom

Comparisons of the pressure catheter derived PWVs at the four levels of afterload to those derived from 2D PC and the real-time RTP (averaged over 10 seconds) scans are shown in the left and right hand plots in Figure 3 respectively. An average catheter derived PWV of 6.35 ± 0.18 m/s was observed, and good agreement was seen from both the 2D PC (6.33 ± 0.70 m/s), and RTP (6.20 ± 0.61 m/s) MR scans.

2D PC vs. RTP Scans in the Volunteers

Figure 4 shows an example M-mode image (right) of the RTP scan at the upper slice level (left), as well as the velocity profiles at the ascending and upper descending aorta after being averaged between the AP kernels. A comparison of post processed velocity data from one cardiac cycle using both scans is shown in Figure 5. Figure 5 A was derived from using the lumina mask filters upon the averaged 2D PC scan data. The plot of Figure 5B is one of the cardiac cycles derived from using the ID kernels upon the M-mode images of the RTP scan.

Post-processed volunteer pulse wave data from one cardiac cycle at the three aortic locations derived from the 2D PC (left) and RTP real time (right) scans.

Figure 6 presents the correlation, (6A, gradient of 0.95, r² = 0.69), and Bland Altman, (6B, mean difference of -0.23 ± 0.51), comparisons between volunteer PWVs derived from the 2D PC and a single RTP scan. Inter scan variability was assessed between the three repeated scans in the right hand plot where standard deviation is plotted against the mean PWV. The mean standard deviation was 0.39 ± 0.29 m/s, approximately 10% of the PWV in our healthy cohort.

Beat-to-Beat PWV and Variation during Respiratory Manoeuvres

We were able to observe beat-to-beat PWV, and its variations over a range of intra-thoracic pressures. Figure 7 (A) shows beat-to-beat PWV (dotted line) and projected velocity profiles (solid lines) at the ascending and lower descending aorta of one volunteer during normal breath hold (top), and Valsalva (middle) and Mueller (bottom) manoeuvres. Additionally, 10 second mean PWV measurements over the three successive measurements of three of the six volunteers (Fig 7B) shows that we were able to observe clear increases in PWV from a normal breath hold to a Mueller manoeuvre in those volunteers.

Discussion

There is clear benefit if an individual assessment of cardiovascular risk can be achieved independent of the subject's physiology during clinical exam. PWV variability has been shown to be reduced in hypertensives [5], therefore we propose that such an individual assessment may become possible by measuring PWV variations induced by modulated intrathoracic pressure. Presented herein is a study of MR derived PWV using a new real time velocity projection MR scan, and its comparisons with pressure catheter derived PWV, standard 2D PC MR, as well as the effects of respiratory manoeuvres. Different techniques have been proposed to determine the PWV from velocity encoded MR data [12, 19-21]. In this embodiment we measured the pulse transit time using the foot-to-foot transit-time method [6] due to its reported reliability in repeated measurements [20], For this we defined the foot as the intersection of the maximum gradient of the upslope according to a moving average window (n = 4), and a horizontal projection from the local mean (n = 5) about the end diastolic minimum.

In- Vitro Performance of RTP for PWV Measurement

The phantom experiments offered a repeatable cardiac cycle with which to assess PWV through the gold standard dual pressure catheter measurements, which was then used as a point of comparison against RTP and current standard 2D phase-encoded MR protocols. According to the Moens-Korteweg equation, PWV is dependent on the Young's modulus, tube radius, tube wall thickness and fluid density [22]. As the afterload was changed, the increased aortic pressure varied the tube dimensions slightly, therefore the gold-standard pressure catheter measurements too varied slightly. The 2D PC and RTP scan protocols incurred PWV errors of the 0.02 and 0.15 m/s respectively, these errors within the 0.1 ±0.8 m/s range found by Westenberg et al, [9]. It is evident therefore that the real time RTP scan is as accurate as 2D PC MRI in assessing PWV when compared to the gold-standard pressure catheter measurement.

We assess PWV in the present embodiment using the foot-to-foot method upon velocity waveform data with a temporal resolution of 11 ms. Accuracy was maintained using this method as the definition of the wave foot is the intersection of two projections according to the data gradient and local minima, rather than a specific data point.

In- Vivo Validation of RTP

A good agreement was observed between the 2D PC and real time RTP MRI scans in vivo as shown in Figure 6. Although neither MR based PWV assessment techniques are considered a gold- standard, this result provides a comparison to the current MR based technique used clinically. The RTP scan showed good repeatability with an inter-scan standard deviation of 0.43 ± 0.29 m/s.

RTP During Respiratory Manoeuvres Figure 7 (A) shows beat-to-beat PWV measurements during normal breath hold, and Valsalva and Mueller manoeuvres in one volunteer. Mean PWV of three volunteers during the manoeuvres is shown in Figure 7 (B) agrees with observations by Hardy et. al [14], that PWV is affected. Most encouraging from this data was the observed beat-to-beat increase in PWV when a Mueller manoeuvre increased the aortic trans-mural stress making the aorta effectively stiffen It is therefore possible to induce aortic PWV changes and observe its behaviour beat-to-beat by assessing transit times with the RTP scan in conjunction with accurate aorta length measurement. The significance of such a technique is that a patent specific relationship between the artery stiffness (i.e. PWV) and the arterial hoop stress (i.e. aortic trans-mural pressure) can be evaluated. We believe that this relationship could have significant diagnostic value as it is independent of arterial pressure.

When aortic trans-mural pressure stress was decreased through a Valsalva manoeuvre, we were not able to see an obvious trend in the effect upon PWV as observed by Latham et. al

[14]_· The puzzling result has been attributed to the complex relationship between respiratory pressure, induced changes in haemodynamics and PWV, which are particularly notable during Valsalva [ 18]. The effect upon PWV during Valsalva is subject to further investigation.

Data from three of the six volunteers asked to perform the respiratory manoeuvres were omitted either due to minimal observed variation in PWV, or due to poor signal preventing PWV analysis. We attribute minimal observed PWV variation being due to lack of instruction in how to carry out a Mueller manoeuvre. Poor signal during one of the manoeuvres was probably due to the physical movement of vessel which then projected velocity into the aorta signal. Abdominal and chest muscle contractions during Valsalva, and/or diaphragm descent during Mueller caused translation of the liver and pulmonary vessels which were both found to affect aortic velocity signal. Practice of the manoeuvres before scanning, as well as careful planning of the RTP slice selections in the latter of the six volunteers improved signal throughout the three manoeuvres. It was found that laying the top slice below the pulmonary artery, and positioning the lower slice 10-20mm above the liver (see Figure 4) were the optimal slice locations.

A suitable breathing protocol is required to maximise the observability of PWV variation. This may include the magnitude of the maintained intra-thoracic pressure, the duration of the manoeuvre, the induced translation of the liver and pulmonary arteries, and even the sequence of manoeuvres. Such a protocol needs to permit high variations in arterial wall strain while maintaining clear signal and should consider the time response of any sympathetic nervous responses affecting blood pressure. Additionally, a means of biofeedback to the subject may be required in the scanner to allow adherence to the protocol. Opportunity to improve the reliability of the RTP scan further exists primarily in the increase of temporal resolution through the simultaneous excitation and read out of both projected velocity profiles. Furthermore, using dual RF coils offset in the feet-head direction so as the lower slice can be moved below the liver to the gut, thus increasing transit time and aortic length. Both of these measures would considerably increase the observability of the velocity waveform permitting higher certainty in individual PWV measurements.

Post-processing of the 1-D M-mode RTP images was less time consuming than the standard 2D PC image data sets. It is relatively simple to select the arterial lumina along only one dimension (see Figure 1), as opposed to generating circular image masks to isolate the flow- encoded signal within the aorta sections. It is also computationally less demanding to process the signal M-mode image where each column of pixels represents a projected time step of 10 seconds, as opposed to the 2D PC data which is a set of 100 phase contrast images for each of the 100 time steps through one single cardiac cycle.

Limitations

Depending on the slice orientation the influence of projected blood velocity from other vessels can be considerable. As described earlier, the lower slice was positioned just above the LV apex. We observed that as this slice is moved down and the liver enters the slice, the amplitude of the velocity waveform is rapidly attenuated. This was attributed to saturating signal from the portal and hepatic circulation. Similarly, the position of the upper slice was affected in some volunteers by blood velocity projected from the pulmonary arteries. This was assumed to be from secondary flow paths after ejection from the pulmonary artery, as moving the upper slice to an oblique axial orientation so as to lie above, or below the pulmonary artery would remove this signal dilution.

Summary

Accuracy of PWV measurements using real time MR acquisitions of velocity projections was shown to be high through repeated experiments. A volunteer study showed that assessed PWV was comparable with current gated, averaged 2D PC, and showed inter-scan variability of 0.39 ± 0.29 m/s. Beat-to-beat PWV assessment was achieved which has not been done using MR to date. Variations in PWV were observable when comparing a normal breath hold and Miieller manoeuvre. MR offers accurate aortic length measurement, and now the ability to observe beat-to-beat PWV.

References

1. MacMahon, S., M.H. Alderman, L.H. Lindholm, L. Liu, R.A. Sanchez and Y.K. Seedat: Blood-Pressure-Related Disease is a Global Healthy Priority. Lancet, 2006. 371: p. 1480-1482.

2. Payne, R.A. and D.J. Webb: Blood Pressure and Stiffness is Hypertension: Is Arterial Structure Important? Hypertension, 2006. 48: p. 366-367.

3. Meaume, S., A. Benetos, O.F. Henry, A. Rudnichi and M. Safar: Aortic Pulse Wave Velocity Predicts Cardiovascular Mortality in Subjects >70 Years of Age. Arterioscler Thromb Vase Biol, 2001. 21: p. 2046-2050.

4. Elias, M.F., G.A. Dore, A. Davey, W.P. Abhayaratna, A.L. Goodell and M.A. Robbins: Norms and Reference Values for Pulse Wave Velocity One Size Does Not Fit

All. THe Journal of Bioscience and Medicine, 2011. 1(4): p. 1-10.

5. Stewart, A.D., B. Jiang, S.C. Millasseau, J.M. Ritter and P.J. Chowienczyk: Acute Reduction of Blood Pressure by Nitroglycerin Does Not Normalize Large Artery Stiffness in Essential Hypertension. Hypertension, 2006. 48: p. 404-410.

6. Laurent, S., J. Cockcroft, L. Van Bortel, et al : Expert consensus document on arterial stiffness: methodological issues and clinical applications. European Heart Journal, 2006. 27: p. 2588-2605.

7. Boese, J.M., M. Bock, S O. Schoenberg and L.R. Schad: Estimation of aortic compliance using magnetic resonance pulse wave velocity measurement. Phys. Med. Biol, 2000. 45: p. 1703-1713.

8. Lotz, J., C. Meier, A. Leppert and M. Galanski: Cardiovascular Flow Measurement with Phase-Contrast MR Imaging: Basic Facts and Implementation. Radiographics, 2002. 22: p. 651-671.

9. Westenberg, J., A. de Roos, H. Grotenhuis, et al : Improved aortic pulse wave velocity assessment from multislice two-directional in-plane velocity-encoded magnetic resonance imaging. J Magn Re son Imaging, 2010. 32(5): p. 1086-1094.

10. Dogui, A., N. Kachenoura, F. Frouin, et al : Consistency of aortic distensibility and pulse wave velocity estimates with respect to the Bramwell-Hill theoretical model: a cardiovascular magnetic resonance study. Journal of Cardiovascular Magnetic Resonance, 2011. 13(11).

11. Hardy, C, B. Bolster, E. McVeigh, W. Adams and E. Zerhouni: A one-dimensional velocity technique for NMR measurement of aortic distensibility. Magn Reson Med, 1994. 31(5): p. 513-520. 12. Taviani, V., A.J. Patterson, M.J. Graves, et al : Accuracy and repeatability of fourier velocity encoded M-mode and two-dimensional cine phase contrast for pulse wave velocity measurement in the descending aorta. J Magn Re son Imaging, 2010. 31(5): p. 1 185-1 194.

13. Hardy, C.J., B.D.J. Bolster, E.R. McVeigh, I.E.T. Iben and E.A. Zerhouni: Pencil Excitation with Interleaved Fourier Velocity Encoding: NMR Measurement of Aortic Distensibility. Magn Resort Med, 1996. 35(6): p. 814-819.

14. Latham, R.D., N. Westerhof, P. Sipkema, B.J. Rubal, P. Reuderink and J.P. Murgo: Regional wave travel and reflections along the human aorta. Circulation, 1985. 72(6): p. 1257-69.

15. Gribbin, B., A. Steptoe and P. Sleight: Pulse Wave Velocity as a Measure of Blood Pressure Change. Psychophysiology, 1976. 13(1): p. 86-90.

16. Naka, K.K., A.C. Tweddel, D. Parthimos, A. Henderson, J. Goodfellow and M.P. Frenneaux: Arterial Distensibility: Acute Changes Following Dynamic Exercise in Normal Subjects. Am J Physiol Heart Circ Physiol, 2002. 248: p. H970-H978.

17. Condos, W.R., R.D. Latham, S.D. Hoadley and P. A. : Hemodynamics of the Miieller Maneuver in Man: Right and Left Heart Micromanometry and Doppler Echocardiography. Circulation, 1987. 76: p. 1020-1028.

18. Korner, P L, A.M. Tonkin and J.B. Uther: Reflex and Mechanical Circulatory Effects of Graded Valsalva Maneuvers in Normal Man. Journal of Applied Physiology, 1976. 40(3): p. 434-440.

19. Dogui, A., A. Redheuil, M. Lefort, et al : Measurement of aortic arch pulse wave velocity in cardiovascular MR: Comparison of transit time estimators and description of a new approach. J Magn Re son Imaging, 201 1. 33(6): p. 1321-1329.

20. Chiu, Y.C., P.W. Arand, S.G. Shroff, T. Feldman and J.D. Carroll: Determination of pulse wave velocities with computerized algorithms. Americian Heart Journal, 1991. 121(5): p. 1460-1470.

21. Ibrahim, E.-S.H., K.R. Johnson, A.B. Miller, J.M. Shaffer and R.D. White: Measuring aortic pulse wave velocity using high-field cardiovascular magnetic resonance: comparison of techniques. Journal of Cardiovascular Magnetic Resonance, 2010. 12(26).

22. Bramwell, J.C. and A.V. Hill. The Velocity of the Pulse Wave in Man. in Royal Society of London. Series B, Containing Papers of a Biological Character . 1922: The Royal Society

Claims

1. A method for characterising aortic stiffness, comprising:

acquiring plural sets of aortic blood-flow velocity data, the sets being acquired under different aortic trans-mural pressures induced by respiratory manoeuvres, the sets comprising blood-flow velocity information from a plurality of aortic locations;

calculating a pulse wave velocity from the sets of aortic blood flow velocity data; and plotting pulse wave velocity against aortic trans-mural pressure;

wherein aortic stiffness may be characterised by the rate of change of pulse wave velocity with aortic trans-mural pressure.

2. A method according to claim 1, and further comprising fitting a predefined function to the plot of pulse wave velocity against aortic trans-mural pressure in order to characterise the relationship therebetween.

3. A method according to claim 1 or 2, and further comprising plotting the pulse wave velocity data against aortic trans-mural pressure on a line graph to produce a graphical image.

4. A method according to any of the preceding claims, wherein the aortic locations comprise two or more selected from the group comprising: the ascending aorta, aortic arch, supra-aortic branches and descending aorta until the diaphragm.

5. A method according to any of the preceding claims, wherein the blood-flow velocity data is acquired using one or more of: an ultra-sound scanner, or a magnetic resonance imaging scanner.

6. A method according to any of the preceding claims, wherein the respiratory manoeuvres include one or more selected from the group comprising: free-breathing, a normal breath-hold, a Valsalva manoeuvre, and a Mueller manoeuvre.

7. A method according to any of the preceding claims, wherein pulse wave velocity is calculated for a set of aortic blood flow velocity data by:

i) processing the set of aortic blood flow velocity data to obtain pulse velocity profiles of blood flow at the aortic locations; ii) determining a transit time between two or more blood velocity waveforms;

iii) measuring distance between the aortic locations; and

iv) calculating pulse wave velocity in dependence on the determined transit time and the measured distance.

8. A method according to claim 7, wherein determination of the transit times uses any of the following:

a) finding the foot of each waveform and measuring time between found feet; or

a. complete waveforms;

or

9. A method for determining pulse wave velocity in a human aorta, comprising:

determining a pulse velocity profile per aortic location;

determining transit times between pulses in the pulse velocity profiles; and

10. A method according to claim 9, and further comprising storing the one dimensional velocity profiles as lines in an M-mode image; and processing the M-mode images to determine the pulse velocity profiles.

11. A method according to any of claims 9 or 10, wherein the scanning comprises any one selected from the group comprising:

iv) scanning using bipolar phase encoding gradients, a sign of which is toggled between successive scans of a slice;

v) scanning using a zero then non-zero phase encoding gradient which are then toggled between successive scans of a slice; or

vi) scanning using either a reference and then successive bipolar or zero/non-zero phase encoding gradients which are subtracted from the reference image.

12. A method according to any of claims 9 to 11, wherein two thoracic transverse slices are scanned, an M-mode image being produced per slice, a one dimensional velocity profile being stored in the M mode image relating to the slice in respect of which the velocity profile was found.

13. A method according to any of claims 8 to 12, wherein two thoracic transverse slices are scanned, the scanning of the two slices being time interleaved, wherein a first slice is scanned using two phase encoding modes, followed by the second slice being scanned using two phase encoding modes.

14. A method according to any of claims 9 to 13, wherein two thoracic transverse slices are scanned, the scanning of each slice being performed in parallel, and determination and storage of one -dimensional velocity profiles also being performed in parallel.

15. A method according to any of claims 9 to 14, wherein the determination of velocity profiles is performed substantially in real-time with the scans.

16. A method according to any of claims 9 to 15, wherein determination of the transit times uses any of the following:

d) finding the foot of each waveform and measuring time between found feet; or e) cross correlating one of the following, and determining the transit time according to a correlation maximum:

d. complete waveforms;

e. data waveform up-slopes determined between the limits of a velocity data between a located velocity toward end diastole, and a located velocity toward the peak of systole; or

f. exponential or sigmoid profiles fitted to waveform up-slopes data determined between the limits of a velocity data between a located velocity toward end diastole, and a located velocity toward the peak of systole;

or

f) determining the transit time according to a minimum of absolute/squared errors between a portion, or all of the velocity waveforms.

17. A method according to any of claims 9 to 16, wherein the subtracting further comprises subtracting a phase encoded velocity projection obtained from a most recent scan of a slice from a phase encoded velocity projection obtained from a scan of the same slice immediately preceding in time the most recent scan, wherein the subtraction is repeated from scan to scan such that a one -dimensional velocity profile is obtained per subsequent scan of the same slice.

18. A method of characterising aortic stiffness according to any of claims 1 to 8, wherein pulse wave velocity is found using a method according to any of claims 9 to 17.

19. An MRI scanner arranged to operate to find aortic pulse wave velocity according to the method of any of claims 9 to 17.

20. An MRI scanner, comprising:

a processor; and

a computer readable medium;

the computer readable medium storing one or more programs so arranged such that when executed by the processor they cause the processor to control the MRI scanner to:

perform magnetic resonance scanning upon one or more thoracic transverse slices positioned so as to intersect at least two aortic locations; subtract phase encoded projections from successive scans of the same slice to obtain one-dimensional velocity profiles;

determine a pulse velocity profile per aortic location;

determine transit times between pulses in the pulse velocity profiles; and

calculate pulse wave velocity in dependence on the transit times and aortic length between the aortic locations.

21. A computer, comprising:

a processor; and

a computer readable medium;

the computer readable medium storing one or more programs so arranged such that when executed by the processor they cause the processor to control the computer to perform the method of any claims 1 to 8.