CN103284702A - Electrocardiogram and pulse wave relation analysis method and method and device of fusion analysis - Google Patents

Abstract

The invention relates to the technical field of measurement for diagnosis, and provides an electrocardiogram and pulse wave relation analysis method and a method and device of fusion analysis based on the electrocardiogram and pulse wave relation analysis method. The electrocardiogram and pulse wave relation analysis method includes the steps that step1, obtained electrocardiogram signals and pulse wave signals of a known patient are preprocessed respectively, and the preprocess comprises base line horizontal-moving and denoising; step2, the preprocessed electrocardiogram and the pulse wave are visually showed in a two-dimensional and/or three-dimensional image; step3, characteristics of the electrocardiogram and the pulse wave are extracted; step4, the characteristics of the electrocardiogram and the pulse wave are compared and analyzed to obtain the mutual relation of the characteristics of the electrocardiogram and the pulse wave, and fusion analysis inspection is carried out according to the mutual relation of the characteristics. Compared with analysis of singular signals, the accuracy rate of classification of results is improved.

Description

Method and device for relationship analysis and fusion analysis of electrocardiogram and pulse wave

【Technical field】

The invention relates to the technical field of measurement for diagnostic purposes, in particular to a method and device for analyzing the relationship between an electrocardiogram and a pulse wave and based on the fusion analysis.

【Background technique】

An electrocardiogram is a graph that records the electrical signals generated by the heart cycle and draws them in a prescribed format. The pulse wave is formed by the pulse of the heart propagating along the arterial vessels and blood flow to the periphery. Although the forms produced by the two are different, the source of both is the heart, so it is of great significance to study the mutually exclusive, consistent and complementary relationship between the two.

ECG has been widely used in the clinical diagnosis of cardiovascular diseases. In recent years, in order to reduce the workload of doctors in remote monitoring and medical examination centers, the demand for automatic ECG analysis has become more and more urgent. However, the classification accuracy of existing ECG classification methods in practical applications is still not satisfactory. In addition to the problems of the ECG classification algorithm itself, the amount of information contained in a single ECG signal is also limited. Therefore, more and more attention is paid to the diagnostic analysis of fusion of multiple signals, but the premise of fusion is to understand the relationship between signals in order to give more accurate analysis conclusions.

In the prior art, there is no method and device for analyzing the relationship between the electrocardiogram and the pulse wave, and performing fusion analysis on the sampled data based on the analysis.

In view of this, it is an urgent problem to be solved in this technical field to overcome the defects in the prior art.

【Content of invention】

The technical problem to be solved by the present invention is to provide a method and device for analyzing the relationship between an electrocardiogram and a pulse wave that fuses multiple signals.

The further technical problem to be solved by the present invention is to provide a method and device for fusion analysis of electrocardiogram and pulse wave based on the aforementioned relationship analysis.

The present invention adopts following technical scheme:

A method for analyzing the relationship between electrocardiogram and pulse wave, comprising:

Step S1: Perform preprocessing on the acquired ECG and pulse wave signals of a subject with a known disease, including baseline translation and denoising;

Step S2: performing visual presentation of two-dimensional and/or three-dimensional images on the preprocessed electrocardiogram and pulse wave;

Step S3: extracting ECG features and pulse wave features;

Step S4: Perform comparative analysis on the features of the electrocardiogram and pulse wave signals to obtain the relationship between the features of the electrocardiogram and pulse wave signals.

The present invention also provides a device for analyzing the relationship between electrocardiogram and pulse wave, comprising:

The preprocessing module is used to preprocess the acquired ECG and pulse wave signals, including baseline translation and denoising;

A visualization presentation module, used for visual presentation of two-dimensional and/or three-dimensional images of the preprocessed electrocardiogram and pulse wave;

The electrocardiogram feature extraction module is used to extract the electrocardiogram feature;

The pulse wave feature extraction module is used to extract the pulse wave feature;

The comparative analysis module is used for comparatively analyzing the characteristics of the electrocardiogram and the pulse wave, and obtaining the relationship between the characteristics of the electrocardiogram and the pulse wave signal.

The present invention also provides a kind of electrocardiogram and pulse wave fusion analysis method, comprising:

Step Q1: Decision-making fusion of the acquired ECG and pulse wave signals of an unknown disease subject;

Step Q2: Using a deep learning classifier, using a deep neural network to perform data layer and feature layer fusion on the ECG and pulse wave signals of the unknown disease subject;

Step Q3: Comprehensively analyze the relationship between the features of the electrocardiogram and the pulse wave signal after the fusion of the decision-making layer, the data layer and the feature layer, and obtain the relationship between the features of the electrocardiogram and the pulse wave signal according to the above relationship analysis method relationship to identify and classify.

The present invention also provides an electrocardiogram and pulse wave fusion analysis device, comprising:

The decision-making layer fusion module is used to carry out decision-making layer fusion to the electrocardiogram and pulse wave signal of an unknown disease object acquired;

The data layer and feature layer fusion module includes a deep learning classifier, and the deep learning classifier utilizes a deep neural network to perform data layer and feature layer fusion on the electrocardiogram and pulse wave signal of the unknown disease object;

The comprehensive analysis module is used to comprehensively analyze the relationship between the characteristics of the electrocardiogram and the pulse wave signal after the fusion of the decision-making layer, the data layer and the feature layer, and analyze the relationship between the electrocardiogram and the pulse wave signal obtained by the above-mentioned relationship analysis device. The relationship between features is identified and classified.

Compared with the prior art, the beneficial effect of the present invention is that: the present invention carries out relational analysis to the signal characteristic of electrocardiogram and pulse wave, obtains the interrelationship between its each characteristic, and carries out fusion analysis detection according to mutual relation, and Compared with the analysis of a single signal, the classification accuracy of the results is improved.

【Description of drawings】

Fig. 1 is a flow chart of a method for analyzing the relationship between an electrocardiogram and a pulse wave in Embodiment 1 of the present invention;

Fig. 2 is a structural block diagram of a device for analyzing the relationship between electrocardiogram and pulse wave in Embodiment 2 of the present invention;

Fig. 3 is a flow chart of an electrocardiogram and pulse wave fusion analysis method according to Embodiment 3 of the present invention;

Fig. 4 is a structural block diagram of an electrocardiogram and pulse wave fusion analysis device according to Embodiment 4 of the present invention.

【Detailed ways】

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.

In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not constitute a conflict with each other.

According to the electrocardiogram and pulse wave relationship analysis method proposed by the present invention, the relationship between the obtained electrocardiogram and pulse wave signal features includes mutual exclusion, consistency and complementarity,

The method for electrocardiogram and pulse wave fusion analysis proposed by the present invention is based on the method for analyzing the relationship between electrocardiogram and pulse wave signals of the present invention, and performs electrocardiogram and pulse wave analysis on the basis of determining the exact relationship between electrocardiogram and pulse wave signals of known disease objects. Wave fusion analysis, compared with the analysis of a single signal, improves the classification accuracy of the results.

Example 1

Embodiment 1 of the present invention provides a method for analyzing the relationship between an electrocardiogram and a pulse wave. The electrocardiogram and pulse wave data in this embodiment are the electrocardiogram and pulse wave data obtained for a specific known cardiovascular disease, such as but not limited to atrial fibrillation, which are not listed here. Taking atrial fibrillation as an example, it is first necessary to obtain the pulse wave and electrocardiogram data of atrial fibrillation.

As shown in Figure 1, the method includes the following steps:

Step S1: Perform preprocessing on the acquired ECG and pulse wave signals of a subject with a known disease, including baseline translation and denoising;

Step S2: performing visual presentation of two-dimensional and/or three-dimensional images on the preprocessed electrocardiogram and pulse wave;

Step S3: extracting ECG features and pulse wave features;

Step S4: Perform comparative analysis on the features of the electrocardiogram and pulse wave signals to obtain the relationship between the features of the electrocardiogram and pulse wave signals. For example: in the data of atrial fibrillation, the pulse rate is inconsistent with the heart rate, etc.

The above steps are described in detail below:

In step S1 preprocessing, taking atrial fibrillation as an example, a band-pass filter of 1-30 Hz can be selected to filter the original pulse wave x[n]. The band-pass filter is designed by Matlab, using a 6th-order Chebyshev filter (chebyshev) type I infinite response filter with two pass-band cut-off points of 1Hz and 30Hz respectively.

The electrocardiogram and pulse wave preprocessed in step S1 are visualized in step S2 and displayed in two-dimensional or three-dimensional images to determine the accuracy of the acquisition position, that is, whether to obtain valid waveform signals. In addition, the relationship between the electrocardiogram and the pulse wave among the intuitive features can also be found out. Taking atrial fibrillation as an example, the 12-dot matrix pulse wave waveform is obtained to determine the accuracy of the acquisition position. According to the characteristics of the pulse wave fluctuating greatly in the center of the vessel, the detection position is judged by the strength of the amplitude fluctuation, so as to ensure the signal quality.

Step S3 can extract electrocardiogram feature and pulse wave feature by following method:

Step S3a: performing electrocardiogram R-wave detection, and extracting the R-wave position in the electrocardiogram;

Step S3b: Perform other time-domain feature detection in the electrocardiogram except the R wave, and extract other time-domain features in the electrocardiogram except the R wave, including but not limited to the extraction of QRS waves;

Step S3c: detecting the main wave of the pulse wave, and extracting the position of the main wave in the pulse wave;

Step S3d: Perform pulse wave time-domain feature detection, including: based on the position of the main wave, detect the starting point of the pulse wave cycle, determine the pulse wave cycle, detect whether the pulse wave cycle is a noise cycle, and detect whether the pulse wave cycle is a non-noise cycle. wave, double wave position amplitude and main wave and cycle form.

The following method can be used to detect whether the pulse wave cycle is a noise cycle: obtain the average pulse wave cycle; based on the average pulse wave cycle, calculate the cosine similarity of each interpolated pulse wave cycle, which is similar to the average pulse wave cycle The pulse wave period whose degree value is less than a set value S is a noise period and is directly excluded. The setting value S can be set according to actual conditions, for example, S=0.8.

Among them, the non-noise periodic detection of the dichroic front wave, the position amplitude of the dichroic wave, the main wave and the periodic shape can be carried out according to the doctor's empirical rules, which can improve the accuracy of time-domain feature extraction, especially the dichroic front wave. According to the experience of TCM doctors and the conclusions of TCM pulse research, the meaning of about 0.05 seconds before the end of the pulse wave is not clear. In addition, if the main wave is detected from the position of the main wave, the main wave may be misdetected as other feature points, so this Detect other time-domain feature points from t seconds after the main wave (for example, t=0.03). Finally determine the detection interval t1 of the double wave and the heavy wave front wave. For example, t1=[p[m]+0.03:pend[m]-0.05] can be set as the time domain feature detection range, where p[m] represents the mth Main wave position, pend[m] indicates the position of the end point of the corresponding pulse wave cycle. According to the statistics, it is found that if the dicrotic front wave exists, the range of occurrence is mostly within the first 1/3 of the range of t1, so the detection range of the dicrotic front wave is t2, and the specific identification method is to find the peak and trough based on the primary and secondary differential signals . This method does not involve complicated steps, and only needs to extract time-domain feature points on the same data set, and compare the extracted results with the labeled results, with high accuracy. The accuracy rates of three obvious time-domain features of main wave, dicrotic wave and dicrotic wave are 99.71%, 88.68%, and 89.03%, respectively, which are higher than other methods.

Specifically, the following methods can be used to detect the R wave of the electrocardiogram:

1a) Band-pass filtering: use a 5-18Hz forward filter to filter the input ECG signal and perform phase delay compensation;

1b) Difference: Perform differential processing on the signal output by the forward filter to form a differential signal;

1c) Data sorting: transform the differential signal: output = absolute value of input/G1, G1 is the assumed maximum value of human R-wave differential value, and can take a value of 0.1 to 0.5. If the output is equal to 1, set it to 1. If the output It is also set to 1 if it is less than 0.01; the data sorting process does not use the normalization method, but uses an absolute value, so that the signal amplitude can be calculated in the subsequent detection, so as to detect the arrest;

2a) Use the formula d(n)*d(n)*log(d(n)*d(n)) to perform Shannon energy conversion on the signal after data arrangement;

2b) Average filtering: Use a forward filter with M=55~75 points (153~208ms) for filtering and phase delay compensation;

3a) Detection of maximum/small points: a maximum point refers to a point whose value is greater than the left and greater than the right, and a minimum point refers to a point whose value is smaller than the left and smaller than the right;

3b) Exclude false R points: Assuming that point n is a maximum point, it will be excluded as long as it meets any of the following conditions:

(a) If the differential signal is less than G2a within the pause time limit of point n, it should be pause here, the maximum point corresponding to point n is not R wave, and G2a is the minimum allowable effective value after R wave difference, which is desirable The value is 0.01~0.06, and its value is related to the signal noise;

(b) If the data value of point n after average filtering is less than G2b, then exclude this point. G2b is the minimum peak value of the allowed R wave, which can be 0.005-0.01;

(c) If the difference between point n and the minimum point next to it is smaller than G3 times the maximum value within the arrest time limit of point n, it will be excluded. G3 is the maximum ratio that allows the amplitude of R wave to suddenly decrease, and it can be 0.06~ 0.2;

3c) Correct mis-excluded points: This operation can correct the situation where the R wave becomes smaller suddenly, and further improve the accuracy rate. The process is as follows: Assume that n points are the maximum points to be excluded, and if all of them meet the following conditions, it is considered to be mis-excluded R wave:

(a) Point n is between two R waves, and the time interval from point n to the two preceding R waves is greater than 2/3 of the time interval between the preceding two R waves;

(b) There are no multiple invalid maximum points between point n and the previous R wave;

(c) The interval from point n to the previous point R and to the next point R is greater than 1/3 second;

(d) On the differential signal, the difference between the amplitude of point n and the minimum point next to it is greater than 0.1 times the difference between the amplitude of the R wave before point n and the minimum point next to it;

(e) On the differential signal, the amplitude difference between the n-point amplitude and the adjacent minimum point is greater than 0.1 times the amplitude difference between the R wave after n point and the adjacent minimum point;

4a) Find the true R position within ±25 points around the approximate R wave position.

Specifically, the following methods can be used to extract other time-domain features of the ECG:

Calculation of time-domain features, respectively calculating the QT interval, the slope of the QRS wave, the slope of the ST segment, and the interval between two adjacent QRS waves;

Form extraction preprocessing, respectively calculate all inflection points for the sampling data in the [start, end] range of P wave, QRS wave and T wave, and further calculate all trend turning points, and locate trend valleys and trend tops; trend valleys refer to those trends The turning point that changes from bottom to top, and the top of the trend refers to the turning point of those trends that change from top to bottom;

Calculation of QRS wave morphological feature points, according to the trend turning point obtained in the preprocessing step, coordinate with the baseline to locate Q wave, R wave, S wave, R' wave and S' wave, and calculate the above-mentioned Q wave, R wave, S wave, R wave at the same time 'wave and S' wave amplitudes;

Morphological recognition, according to the results of the shape extraction preprocessing step to obtain the morphological features of the P wave and T wave; according to the calculation results of the QRS wave morphological feature point calculation step, identify the QRS wave morphological pattern, and obtain the morphological features of the QRS wave.

Specifically, pulse wave main wave detection can be performed by the following methods:

1) The pulse wave is first filtered by a 1Hz-30Hz band-pass filter to remove low-frequency and high-frequency noise;

2), and then perform amplitude normalization on the filtered signal;

3) Calculate the Shannon energy after normalizing the signal amplitude, and perform low-pass filtering to obtain the envelope;

4) Hilbert transform the Shannon energy envelope signal;

5) Use the Hilbert transformed signal to extract the peak point t of the Shannon energy envelope, that is, the value of the Hilbert transformed signal from negative to positive zero crossing corresponds to the peak point of the Shannon energy envelope;

6) Determine the real main wave position in the original pulse wave signal, and determine the time range T centered on the time position t of the Shannon energy peak point, and find the real main pulse wave position in the T range of the original pulse wave signal; 0.2<T< 0.3.

Of course, other known methods for detecting the R wave of the electrocardiogram, extracting other time-domain features, and detecting the main wave of the pulse wave can also be used. The above description is only for illustration, and the present invention is not limited thereto.

Preferably, when extracting the features of the electrocardiogram and the pulse wave, the frequency domain features of the electrocardiogram and the pulse wave can also be extracted by adopting a frequency domain feature analysis method. Frequency domain feature analysis methods include but are not limited to frequency domain feature analysis methods such as Fourier transform, wavelet transform, wavelet packet, and high-order statistics, and the method selection is determined according to the specific situation.

In the comparative analysis of step S4, taking atrial fibrillation as an example, in a specific embodiment, the electrocardiogram conforms to the characteristics of atrial flutter and atrial fibrillation, and the heart rate (RR interval) and pulse rate (pulse wave main wave interval) are calculated, and it is found that In atrial fibrillation the pulse rate does not coincide with the heart rate, whereas in atrial flutter it does.

Example 2

Embodiment 2 of the present invention provides a device for analyzing the relationship between electrocardiogram and pulse wave. The device adopts the method provided in embodiment 1 to analyze the relationship between electrocardiogram and pulse wave. The electrocardiogram and pulse wave data in this embodiment are the electrocardiogram and pulse wave data obtained for a specific known cardiovascular disease, such as but not limited to atrial fibrillation, which are not listed here.

As shown in FIG. 2 , the device includes a preprocessing module 1 , a visual presentation module 2 , an electrocardiogram feature extraction module 3 , a pulse wave feature extraction module 4 and a comparative analysis module 6 . Among them, the preprocessing module 1 performs preprocessing on the acquired ECG and pulse wave signals respectively, including baseline translation and denoising; the visualization presentation module 2 performs two-dimensional and/or three-dimensional visual presentation of the preprocessed ECG and pulse wave signals The electrocardiogram feature extraction module 3 extracts the electrocardiogram feature; the pulse wave feature extraction module 4 extracts the pulse wave feature; the comparative analysis module 6 compares and analyzes the electrocardiogram feature and the pulse wave feature, and draws the relationship between each feature of the electrocardiogram and the pulse wave signal.

ECG features mainly include R wave and other time domain features of ECG, pulse wave features mainly include main wave and other time domain features of pulse wave, therefore, ECG feature extraction module 3 can further include ECG R wave detection unit 3a and other time domain features of ECG Feature extraction unit 3b, wherein, electrocardiogram R wave detection unit 3a detects electrocardiogram R wave, extracts R wave position in electrocardiogram; other time domain feature extraction unit 3b of electrocardiogram performs other time domain feature detection except R wave in electrocardiogram, extracts electrocardiogram other temporal features in addition to the R wave.

The pulse wave feature extraction module 4 can further include a pulse wave main wave detection unit 4a and a pulse wave time-domain feature extraction unit 4b, wherein the pulse wave main wave detection unit 4a detects the pulse wave main wave and extracts the main wave position in the pulse wave The pulse wave time-domain feature extraction unit 4b carries out pulse wave time-domain feature detection, including: taking the position of the main wave as the basis, carrying out pulse wave cycle starting point detection, determining the pulse wave cycle, and detecting whether the pulse wave cycle is a noise cycle, for non The noise cycle detects the front wave of the dichroic wave, the position and amplitude of the dichotomic wave, the main wave and the periodic shape.

Preferably, the device may further include a frequency-domain feature extraction module 5, which extracts the frequency-domain features of the electrocardiogram and pulse wave through a frequency-domain feature analysis method.

Example 3

Embodiment 3 of the present invention provides a method for analyzing the relationship between electrocardiogram and pulse wave. The electrocardiogram and pulse wave data that need to be processed in this embodiment are the electrocardiogram and pulse wave data obtained for a subject with an unknown disease. enumerate. The implementation of this method relies on Example 1, that is, it needs to be carried out on the basis of Example 1.

As shown in Figure 3, the method includes the following steps:

Step Q1: Decision-making fusion of the acquired ECG and pulse wave signals of an unknown disease subject;

Step Q2: Using a deep learning classifier, using a deep neural network to perform data layer and feature layer fusion on the ECG and pulse wave signals of unknown disease subjects;

Step Q3: Comprehensively analyze the relationship between the features of the electrocardiogram and the pulse wave signal after the fusion of the decision-making layer, the data layer and the feature layer, and perform a comprehensive analysis according to the relationship between the features of the electrocardiogram and the pulse wave signal obtained in Example 1 Identify categories.

The above steps are described in detail below:

Step Q1 specifically includes:

Step Q1a: performing electrocardiogram R-wave detection, and extracting the R-wave position in the electrocardiogram;

Step Q1b: Perform other time-domain feature detection in the electrocardiogram except the R wave, and extract other time-domain features in the electrocardiogram except the R wave;

Step Q1c: Perform pulse wave main wave detection, and extract the main wave position in the pulse wave;

Step Q1d: Perform pulse wave time-domain feature detection, including: based on the position of the main wave, detect the starting point of the pulse wave cycle, determine the pulse wave cycle, detect whether the pulse wave cycle is a noise cycle, and detect whether the pulse wave cycle is a non-noise cycle. wave, double wave position amplitude and main wave and cycle form;

Step Q1e: According to the relationship between the features of the electrocardiogram and the pulse wave signal obtained in embodiment 1, rule-based reasoning is performed on the relationship between the features of the electrocardiogram and the pulse wave signal extracted in steps Q1a-Q1d.

Wherein, the methods provided in steps Q1a-Q1d are similar to the methods in steps S3a-S3d in Embodiment 1. The details will not be described in detail here, please refer to Embodiment 1.

In step Q2, the fusion of the data layer and the feature layer can adopt the following method: after determining the position of the main wave of the pulse wave, calculate the direct eigenvalue of the pulse wave feature, and perform convolution and sampling on the segmented data segments to obtain the internal eigenvalue , combined with direct eigenvalues and internal eigenvalues to calculate according to a predetermined algorithm to obtain classification results. Instead of extracting various eigenvalues with low accuracy and being easily disturbed by noise (such as dicrotic front wave), direct eigenvalues with high accuracy (such as main wave interval) are directly extracted and included in the algorithm for calculation , can improve the final classification accuracy and output more accurate pulse wave classification results. It can be understood that other known methods can also be used to fuse the data layer and the feature layer, and the above description is only for illustration, and the present invention is not limited thereto.

Example 4

Embodiment 4 of the present invention provides a device for analyzing the relationship between electrocardiogram and pulse wave. The electrocardiogram and pulse wave data that need to be processed in this embodiment are the electrocardiogram and pulse wave data obtained for a subject with an unknown disease. enumerate. The implementation of the device depends on Embodiment 2, that is, it needs to be performed on the basis of Embodiment 2.

As shown in Figure 4, the device includes a decision-making layer fusion module 7, a data layer and feature layer fusion module 8 and a comprehensive analysis module 9, wherein the decision-making layer fusion module 7 performs an electrocardiogram and a pulse wave signal on an acquired unknown disease object. Decision-making layer fusion; data layer and feature layer fusion module 8 includes a deep learning classifier, and the deep learning classifier utilizes deep neural network to carry out data layer feature layer fusion to the electrocardiogram and pulse wave signal of unknown disease object; comprehensive analysis module 9 pairs of decision-making Layer fusion, the relationship between the electrocardiogram and the pulse wave signal features after the fusion of the data layer and the feature layer is analyzed comprehensively, and the relationship between the electrocardiogram and the pulse wave signal features is identified and classified according to embodiment 2.

Preferably, the decision-making layer fusion module 7 specifically includes an electrocardiogram R wave detection unit 7a, other time-domain feature extraction units 7b of the electrocardiogram, a main pulse wave detection unit 7c, a pulse wave time-domain feature extraction unit 7d and a rule reasoning unit 7e, wherein, The electrocardiogram R wave detection unit 7a detects the electrocardiogram R wave, and extracts the R wave position in the electrocardiogram; the other time domain feature extraction unit 7b of the electrocardiogram performs other time domain feature detection except the R wave in the electrocardiogram, and extracts other time domain features except the R wave in the electrocardiogram. Time-domain features; pulse wave main wave detection unit 7c detects pulse wave main wave and extracts main wave position in pulse wave; pulse wave time-domain feature extraction unit 7d performs pulse wave time-domain feature detection, including: based on main wave position , to detect the starting point of the pulse wave cycle, to determine the pulse wave cycle, to detect whether the pulse wave cycle is a noise cycle, to detect the non-noise cycle of the front wave of the pulse wave, the position amplitude of the pulse wave, the main wave and the cycle shape; the rule reasoning unit 7e according to The relationship between the electrocardiogram obtained in embodiment 2 and each feature of the pulse wave signal, for the electrocardiogram R wave detection unit 7a, other time domain feature extraction units 7b of the electrocardiogram, pulse wave main wave detection unit 7c and pulse wave time domain feature extraction unit 7d The relationship between the extracted ECG features and pulse wave features is used for rule reasoning.

Wherein, the functions of the electrocardiogram R wave detection unit 7a, other time domain feature extraction units 7b of the electrocardiogram, the pulse wave main wave detection unit 7c and the pulse wave time domain feature extraction unit 7d are respectively the same as those of the electrocardiogram R wave detection unit 3a, the pulse wave time domain feature extraction unit 7d in embodiment 2. The other time-domain feature extraction unit 3b of the electrocardiogram, the main pulse wave detection unit 3c and the pulse wave time-domain feature extraction unit 3d are similar. During specific implementation, two functions can be realized by one unit, and two functions can also be realized by two units respectively. The electrocardiogram R wave also detects the R wave of the unknown disease object; two electrocardiogram R wave detection units can also be respectively set in the device of this embodiment: the electrocardiogram R wave detection unit 7a and the electrocardiogram R wave detection unit 3a detect unknown diseases respectively ECG R-waves of subjects and subjects with known disease.

It is worth noting that the information interaction and execution process between the above-mentioned devices and modules and units in the system are based on the same idea as the embodiment of the processing method of the present invention, and the specific content can refer to the description in the embodiment of the method of the present invention , which will not be repeated here.

Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the embodiments can be completed by instructing related hardware through a program. The program can be stored in a computer-readable storage medium, and the storage medium can include: only Read memory (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), disk or CD, etc.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention should be included in the protection of the present invention. within range.

Claims (12)

1. an electrocardiogram and pulse wave relationship analysis method is characterized in that, comprising:

Step S1: electrocardiogram and pulse wave signal to the known disease object that obtains carry out pretreatment respectively, comprise baseline translation and denoising;

Step S2: pretreated electrocardiogram and pulse wave are carried out the visual of two dimension and/or 3-D view present;

Step S3: extract Characteristics of electrocardiogram and pulse wave feature;

Step S4: Characteristics of electrocardiogram and pulse wave feature are analyzed, draw the relation between each feature of electrocardiogram and pulse wave signal.

2. electrocardiogram according to claim 1 and pulse wave relationship analysis method is characterized in that, among the described step S3, the method for extracting Characteristics of electrocardiogram specifically comprises:

Step S3a: carry out ECG R wave and detect, extract R ripple position in the electrocardiogram;

Step S3b: carry out other temporal signatures detections except the R ripple in the electrocardiogram, extract other temporal signatures except the R ripple in the electrocardiogram;

The method of extracting the pulse wave feature specifically comprises:

Step S3c: carry out pulse wave master ripple and detect, extract main ripple position in the pulse wave;

Step S3d: carry out the pulse wave temporal signatures and detect, comprise: be foundation with described main ripple position, carrying out the pulse wave cycle starting point detects, determine pulse wave cycle, whether be noise periods, non-noise periods is detected heavily rich prewave, heavily rich ripple position amplitude and the main cycle form that involves if detecting described pulse wave cycle.

3. electrocardiogram according to claim 2 and pulse wave relationship analysis method is characterized in that, in described step S3d, whether the described pulse wave cycle of described detection is that the step of noise periods is specially:

Obtain average pulse period of wave;

Based on described average pulse period of wave, calculate each through the cosine-shaped similarity of the pulse wave cycle after interpolation, with the similarity value of average pulse period of wave be noise periods less than the pulse wave cycle of a setting value S.

4. electrocardiogram according to claim 2 and pulse wave relationship analysis method is characterized in that, among the described step S3, the method for extracting Characteristics of electrocardiogram and pulse wave feature also comprises:

Adopt the frequency domain character analytical method, extract the frequency domain character of electrocardiogram and pulse wave.

5. an electrocardiogram and pulse wave relationship analysis apparatus is characterized in that, comprising:

Pre-processing module is used for electrocardiogram and the pulse wave signal that obtains carried out pretreatment respectively, comprises baseline translation and denoising;

The visual module that presents is used for that pretreated electrocardiogram and pulse wave are carried out the visual of two dimension and/or 3-D view and presents;

The Characteristics of electrocardiogram extraction module is used for extracting Characteristics of electrocardiogram;

The pulse wave characteristic extracting module is used for extracting the pulse wave feature;

The relative analysis module is used for Characteristics of electrocardiogram and pulse wave feature are analyzed, and draws the relation between each feature of electrocardiogram and pulse wave signal.

6. electrocardiogram according to claim 5 and pulse wave relationship analysis apparatus is characterized in that, described Characteristics of electrocardiogram extraction module specifically comprises:

The ECG R wave detecting unit is used for carrying out ECG R wave and detects, and extracts R ripple position in the electrocardiogram;

Other temporal signatures extraction units of electrocardiogram are used for carrying out electrocardiogram other temporal signatures except the R ripple and detect, and extract other temporal signatures except the R ripple in the electrocardiogram;

Described pulse wave characteristic extracting module specifically comprises:

Pulse wave master ripple detecting unit is used for carrying out pulse wave master ripple and detects, and extracts main ripple position in the pulse wave;

Pulse wave temporal signatures extraction unit, being used for carrying out the pulse wave temporal signatures detects, comprise: be foundation with described main ripple position, carrying out the pulse wave cycle starting point detects, determine pulse wave cycle, whether be noise periods, non-noise periods is detected heavily rich prewave, heavily rich ripple position amplitude and the main cycle form that involves if detecting described pulse wave cycle.

7. electrocardiogram according to claim 6 and pulse wave relationship analysis apparatus is characterized in that, whether described pulse wave temporal signatures extraction unit detects described pulse wave cycle is that the method for noise periods is specially:

Obtain average pulse period of wave;

Based on described average pulse period of wave, calculate each through the cosine-shaped similarity of the pulse wave cycle after interpolation, with the similarity value of average pulse period of wave be noise periods less than the pulse wave cycle of a setting value S.

8. electrocardiogram according to claim 5 and pulse wave relationship analysis apparatus is characterized in that, described device also comprises:

The frequency domain character extraction module is used for extracting the frequency domain character of electrocardiogram and pulse wave by the frequency domain character analytical method.

9. an electrocardiogram and pulse wave convergence analysis method is characterized in that, comprising:

Step Q1: the electrocardiogram of the unidentified illness object that obtains and pulse wave signal are carried out decision-making level merge;

Step Q2: adopt degree of deep learning classification device, utilize the deep layer neutral net to the electrocardiogram of described unidentified illness object and pulse wave signal carries out data Layer and characteristic layer merges;

Step Q3: the electrocardiogram after the fusion of described decision-making level, data Layer and the characteristic layer fusion and the relation between each feature of pulse wave signal are carried out analysis-by-synthesis, and carry out discriminator according to the relation between each electrocardiogram that obtains of claim 1-4 and each feature of pulse wave signal.

10. electrocardiogram according to claim 9 and pulse wave convergence analysis method is characterized in that described step Q1 specifically comprises:

Step Q1a: carry out ECG R wave and detect, extract R ripple position in the electrocardiogram;

Step Q1b: carry out other temporal signatures detections except the R ripple in the electrocardiogram, extract other temporal signatures except the R ripple in the electrocardiogram;

Step Q1c: carry out pulse wave master ripple and detect, extract main ripple position in the pulse wave;

Step Q1d: carry out the pulse wave temporal signatures and detect, comprise: be foundation with described main ripple position, carrying out the pulse wave cycle starting point detects, determine pulse wave cycle, whether be noise periods, non-noise periods is detected heavily rich prewave, heavily rich ripple position amplitude and the main cycle form that involves if detecting described pulse wave cycle;

Step Q1e: according to the relation between each electrocardiogram that obtains of claim 1-4 and each feature of pulse wave signal, the Characteristics of electrocardiogram extracted among the step Q1a-Q1d and the relation between the pulse wave feature are carried out rule-based reasoning.

11. an electrocardiogram and pulse wave convergence analysis device is characterized in that, comprising:

Decision-making level's Fusion Module is used for that the electrocardiogram of the unidentified illness object that obtains and pulse wave signal are carried out decision-making level and merges;

Data Layer and characteristic layer Fusion Module comprise a degree of deep learning classification device, and described degree of deep learning classification device utilizes the deep layer neutral net to the electrocardiogram of described unidentified illness object and pulse wave signal carries out data Layer and characteristic layer merges;

The analysis-by-synthesis module, be used for the electrocardiogram after the fusion of described decision-making level, data Layer and the characteristic layer fusion and the relation between each feature of pulse wave signal are carried out analysis-by-synthesis, and carry out discriminator according to the relation between each electrocardiogram that obtains of claim 5-8 and each feature of pulse wave signal.

12. electrocardiogram according to claim 11 and pulse wave convergence analysis device is characterized in that, described decision-making level Fusion Module specifically comprises:

The ECG R wave detecting unit is used for carrying out ECG R wave and detects, and extracts R ripple position in the electrocardiogram;

Other temporal signatures extraction units of electrocardiogram are used for carrying out electrocardiogram other temporal signatures except the R ripple and detect, and extract other temporal signatures except the R ripple in the electrocardiogram;

Pulse wave master ripple detecting unit is used for carrying out pulse wave master ripple and detects, and extracts main ripple position in the pulse wave;

Pulse wave temporal signatures extraction unit, being used for carrying out the pulse wave temporal signatures detects, comprise: be foundation with described main ripple position, carrying out the pulse wave cycle starting point detects, determine pulse wave cycle, whether be noise periods, non-noise periods is detected heavily rich prewave, heavily rich ripple position amplitude and the main cycle form that involves if detecting described pulse wave cycle;

The rule-based reasoning unit, be used for according to the relation between each electrocardiogram that obtains of claim 5-8 and each feature of pulse wave signal, ECG R wave detecting unit, other temporal signatures extraction units of electrocardiogram, pulse wave master ripple detecting unit and the Characteristics of electrocardiogram of pulse wave temporal signatures extraction unit extraction and the relation between the pulse wave feature are carried out rule-based reasoning.

Images