CN112587119B - Peripheral artery feature extraction method based on deep learning - Google Patents
Peripheral artery feature extraction method based on deep learning Download PDFInfo
- Publication number
- CN112587119B CN112587119B CN202011456466.4A CN202011456466A CN112587119B CN 112587119 B CN112587119 B CN 112587119B CN 202011456466 A CN202011456466 A CN 202011456466A CN 112587119 B CN112587119 B CN 112587119B
- Authority
- CN
- China
- Prior art keywords
- blood flow
- data
- peripheral artery
- deep learning
- network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000013135 deep learning Methods 0.000 title claims abstract description 27
- 210000001367 artery Anatomy 0.000 title claims description 14
- 238000000605 extraction Methods 0.000 title claims description 8
- 230000002093 peripheral effect Effects 0.000 title claims description 6
- 230000017531 blood circulation Effects 0.000 claims abstract description 76
- 238000000034 method Methods 0.000 claims abstract description 44
- 208000030613 peripheral artery disease Diseases 0.000 claims abstract description 35
- 230000008859 change Effects 0.000 claims abstract description 27
- 241000282414 Homo sapiens Species 0.000 claims abstract description 11
- 238000009792 diffusion process Methods 0.000 claims abstract description 11
- 238000012360 testing method Methods 0.000 claims abstract description 9
- 238000005259 measurement Methods 0.000 claims abstract description 8
- 210000001519 tissue Anatomy 0.000 claims description 45
- 238000012549 training Methods 0.000 claims description 33
- 239000000523 sample Substances 0.000 claims description 22
- 238000011176 pooling Methods 0.000 claims description 12
- 238000001514 detection method Methods 0.000 claims description 10
- 238000013528 artificial neural network Methods 0.000 claims description 7
- 230000007246 mechanism Effects 0.000 claims description 7
- 238000007781 pre-processing Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 238000011084 recovery Methods 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 6
- 238000001228 spectrum Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 5
- 210000003141 lower extremity Anatomy 0.000 claims description 4
- 210000002027 skeletal muscle Anatomy 0.000 claims description 4
- 230000035488 systolic blood pressure Effects 0.000 claims description 4
- 230000000903 blocking effect Effects 0.000 claims description 3
- 125000004122 cyclic group Chemical group 0.000 claims description 3
- 210000000689 upper leg Anatomy 0.000 claims description 3
- 102100032202 Cornulin Human genes 0.000 claims description 2
- 101000920981 Homo sapiens Cornulin Proteins 0.000 claims description 2
- 238000005311 autocorrelation function Methods 0.000 claims description 2
- 238000013480 data collection Methods 0.000 claims 2
- 238000003745 diagnosis Methods 0.000 abstract description 25
- 208000005764 Peripheral Arterial Disease Diseases 0.000 abstract description 24
- 238000005516 engineering process Methods 0.000 abstract description 9
- 238000005100 correlation spectroscopy Methods 0.000 abstract description 4
- 230000007547 defect Effects 0.000 abstract description 4
- 239000000284 extract Substances 0.000 abstract description 3
- 208000030831 Peripheral arterial occlusive disease Diseases 0.000 description 8
- 238000010586 diagram Methods 0.000 description 5
- 238000013527 convolutional neural network Methods 0.000 description 4
- 201000010099 disease Diseases 0.000 description 4
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 210000003423 ankle Anatomy 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 230000002526 effect on cardiovascular system Effects 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 238000002329 infrared spectrum Methods 0.000 description 2
- 238000004445 quantitative analysis Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 208000031104 Arterial Occlusive disease Diseases 0.000 description 1
- 208000024172 Cardiovascular disease Diseases 0.000 description 1
- 206010069729 Collateral circulation Diseases 0.000 description 1
- 230000005526 G1 to G0 transition Effects 0.000 description 1
- 206010022562 Intermittent claudication Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 208000021328 arterial occlusion Diseases 0.000 description 1
- 210000004204 blood vessel Anatomy 0.000 description 1
- 210000002302 brachial artery Anatomy 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 208000029078 coronary artery disease Diseases 0.000 description 1
- 210000004351 coronary vessel Anatomy 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 210000003743 erythrocyte Anatomy 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000000004 hemodynamic effect Effects 0.000 description 1
- 208000021156 intermittent vascular claudication Diseases 0.000 description 1
- 238000007917 intracranial administration Methods 0.000 description 1
- 201000002818 limb ischemia Diseases 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 230000000474 nursing effect Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000000284 resting effect Effects 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 238000004611 spectroscopical analysis Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
- 230000009885 systemic effect Effects 0.000 description 1
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
- A61B5/026—Measuring blood flow
- A61B5/0261—Measuring blood flow using optical means, e.g. infrared light
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7271—Specific aspects of physiological measurement analysis
- A61B5/7275—Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Abstract
The invention discloses a peripheral artery disease diagnosis method based on deep learning, which aims at the defects of peripheral artery disease diagnosis based on ABI in the traditional method, obtains the tissue blood flow change data of a peripheral artery disease patient and a healthy volunteer by using a diffusion correlation spectroscopy technology, trains the tissue blood flow data based on the proposed deep learning network, extracts key features containing peripheral artery disease information, trains to obtain a deep learning network model for diagnosing the peripheral artery disease, inputs test set data into a peripheral artery disease diagnosis model for diagnosis, and realizes the diagnosis of the peripheral artery disease. The method not only makes up the defects of the conventional ABI for diagnosing the peripheral artery diseases, but also verifies the feasibility of the tissue blood flow measurement for diagnosing the peripheral artery diseases, and provides a new technology and a new method for diagnosing the peripheral artery diseases.
Description
Technical Field
The invention relates to the technical field of biomedical engineering, in particular to a peripheral artery disease diagnosis method based on deep learning.
Background
Peripheral Arterial Disease (PAD) is a common Disease in cardiovascular clinic, and the main symptom of intermittent claudication is that the Disease rate is more than four times that of coronary heart Disease, and the Disease rate is that the Disease rate is narrow or blocked in systemic arteries except coronary arteries and intracranial arteries [1-2] . The clinical and epidemiological research results show that the incidence rate of the cardiovascular and cerebrovascular complications of the patients with the peripheral artery diseases is 2 to 4 times of that of the same-age people without the peripheral artery diseases [3] . A great deal of research proves that the blood flow of the peripheral artery disease changes earlier than the structure, and the diagnosis and the evaluation of the degree of the blood flow change are particularly important for the quantitative analysis of the peripheral artery disease. The Ankle Brachial Index (ABI) is the ratio of the Ankle arterial systolic pressure to the upper arm arterial systolic pressure and is a criterion for diagnosing and assessing peripheral arterial disease in clinical and epidemiological studies. ABI is closely related to blood flow change, and the smaller the ABI value is, the more serious the artery blockage and limb ischemia degree is [4] . However, ABI, as a diagnostic criterion for peripheral artery disease, has various disadvantages: ABI is only suitable for detection in a resting state and cannot be used under complex actions such as a moving plate and the likeCarrying out detection; when the artery of the lower limb is calcified to cause the blood vessel to be incompressible, or when the brachial artery is severely stenosed but has enough artery collateral circulation, the ABI cannot be measured or the measurement result is inaccurate; ABI is not related to the patient's Walking ability parameters such as Maximum Walking Time (MWT).
Diffusion Correlation Spectroscopy (DCS) [5-7] The method is a new near infrared spectrum technology, can directly obtain the blood flow change of deep tissues, can continuously detect for a long time, and provides possibility for realizing the quantitative analysis of peripheral arterial diseases. The diffusion correlation spectrum irradiates the tissue surface by utilizing near infrared spectrum, and the light intensity autocorrelation function (g) of the scattered light spot on the tissue surface is calculated 2 (tau)) calculating the motion state of the red blood cells in the tissue to obtain a Blood Flow Index (BFI), thereby realizing the quantitative detection of the tissue Blood Flow (BF). The great advantage of the diffusion-related spectroscopy-based technology in deep tissue blood flow detection is that the obtained tissue blood flow change data contains characteristic information for diagnosing peripheral artery diseases. The deep learning method can directly extract required useful features from raw data. In recent years, there has been literature that implements feature extraction of physiological signals based on a classical deep learning framework. Bi Xiaojun and the like [8] The improved Convolutional Neural Network (CNN) and the neural turing machine are combined to build a deep learning mixed model, more electroencephalogram abstract and deep features are extracted in electroencephalogram robust feature learning, and the advantages of the deep learning method in the aspect of physiological signal feature extraction are explained.
The invention is used for measuring the blood flow of gastrocnemius muscle tissues of patients with peripheral artery diseases and healthy people on the basis of detecting the blood flow of deep tissues of a diffusion related spectrum, realizes the diagnosis of the peripheral artery diseases by using a deep learning method on the basis of the obtained tissue blood flow change data, and provides a new technology and a new method for the diagnosis of the peripheral artery diseases and other related cardiovascular diseases.
Reference documents:
[1]Sieminski D J,Gardner A W.The relationship between free-living daily physical activity andthe severity ofperipheral arterial occlusive disease[J].Vascular Medicine,1997,2(4):286-291.
[2]Treatjacobson D,Halverson S L,Ratchford A,et al.A patient-derived perspective of health-related quality of life with peripheral arterial disease[J].Journal of Nursing Scholarship,2002,34(1):55-60.
[3] zhang Xiongwei peripheral arterial disease non-invasive hemodynamic detection technique [ M ] human health press, 2010 [9]
[4]Norgren L,Hiatt WR,Dormandy JA,et al.Inter-society consensus for the management of peripheral arterial disease(TASC II)[J].European Journal of Vascular and Endovascular Surgery,2007,33:S1-S75
[5]Pine D J,Weitz D A,Chaikin P M and Herbolzheimer E.Diffusing-wave spectroscopy[J].Physical Review Letters,1988,60(12):1134-1137.
[6]Maret G,Wolf P E.Multiple light scattering:weak localization and dynamic fluctuations[J].Physica Scripta,1989:223-225.
[7]Boas D A,Campbell L E,Yodh A G,et al.Scattering and imaging with diffusing temporal field correlations[J].Physical Review Letters,1995,75(9):1855-1858.
[8] Bi Xiaojun, qiao Weizheng EEG robust feature learning based on improved deep learning model C-NTM [ J ]. Harbin university of engineering proceedings 2019,40 (9): 1642-1649.
Disclosure of Invention
In order to solve the problems of inaccuracy, manual calculation, time consumption and labor consumption of the traditional ABI detection method, the invention provides a peripheral artery disease diagnosis method based on deep learning. The invention aims to perform characteristic analysis and extraction on the collected tissue blood flow change signals by using a deep learning method, so that the diagnosis of peripheral artery diseases is realized by using the extracted characteristic information. The method not only makes up the defects of the conventional ABI for diagnosing the peripheral artery diseases, but also verifies the feasibility of the tissue blood flow measurement for diagnosing the peripheral artery diseases, and provides a new technology and a new method for diagnosing the peripheral artery diseases.
In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
a peripheral artery disease diagnosis method based on deep learning comprises the following steps:
step (1): and (6) data acquisition.
Collecting tissue blood flow change data of gastrocnemius positions of a tested patient with peripheral artery disease and a healthy volunteer by a diffusion related spectrum blood flow detection system; during measurement, the testees are in a lying state, the probe is placed at the gastrocnemius part of each testee, and the cuff is placed at the root of the thigh of the lower limb measured by the testee and used for blocking blood flow.
Step (2): and (4) preprocessing data.
And (2) after normalization preprocessing is carried out on the tissue blood flow change data obtained in the step (1), randomly dividing the tissue blood flow change data of the tested person into a training set and a testing set. Then, the training set samples are subjected to data enhancement through a data enhancement method, and the training set samples after the data enhancement are used as data input of the deep learning network.
And (3): and (5) constructing a deep learning network.
Aiming at the characteristics of the collected tissue blood flow change data, the invention provides a dual-view convolution cyclic neural network (dual-view CRNN, dvCRNN), wherein a dual-view mechanism is introduced into the network on the basis of the Convolution Neural Network (CNN), so that the characteristic information in a sample can be fully provided; firstly, respectively inputting g2 (tau) into two convolution network layers, respectively carrying out maximum pooling and average pooling after convolution is finished, wherein the maximum pooling is used for extracting and learning detail features of blood flow signals, the average pooling is used for extracting and learning overall features of blood flow, then splicing the features, leveling and inputting spliced data into an RNN network gating cycle unit (GRU), extracting time sequence features of the signals, classifying through a full connection layer, sending a classification result into a softmax network to obtain the probability of whether the classification result is a patient, and using a cross entropy function as a loss function; using Adam as an optimizer to find the optimal solution of the model;
cross entropy loss function:
L＝-[y i ·log(p i )+(1-y i )·log(1-p i )]
wherein: yi represents the label of the sample i, the positive class is 1, and the negative class is 0;
pi represents the probability that sample i is predicted to be positive;
and (4): and (5) network training and diagnosis.
Inputting the enhanced training set sample data in the step (2) into the dvCRNN network constructed in the step (3) for training, and inputting the test set sample data into the trained network model for diagnosis after the training is finished to obtain a diagnosis result.
Further, the device for acquiring the tissue blood flow change data in the step (1) is a tissue blood flow acquisition system based on a diffusion correlation spectroscopy technology.
Further, in the step (1), the tissue blood flow collection process of two groups of testees, namely the peripheral artery disease patient and the healthy volunteer, is divided into three stages. A plateau of 1min, an arterial occlusion phase of 5min, and a recovery phase of 10min, respectively.
Further, the duration of the tissue blood flow change data acquired in the step (1) is 16min in total, and the data length is 150 sample points.
Furthermore, the tissue blood flow data of the patient with peripheral artery disease collected in step (2) totals 166 samples, and the data of healthy volunteers totals 94 samples. Randomly selecting 130 samples from 166 peripheral artery disease patient tissue blood flow samples as training set samples, and taking the rest 36 samples as prediction set samples; similarly, 70 of 94 healthy volunteer data tissue blood flow samples were randomly selected as training set samples, and the remaining 24 were selected as prediction set samples.
Further, the data enhancement algorithm involved in step (2) is a down-sampling-up-sampling method, by which the 130 patients with peripheral artery disease tissue blood flow data samples and 70 healthy volunteers tissue blood flow data samples of the training set are enhanced to 390 cases and 210 cases respectively.
Further, in the dvCRNN network constructed in step (3), the dual-view mechanism is to extract and learn features of the blood flow signal from the detail and whole levels, and obtain the detail and whole features in the signal by respectively adopting maximum pooling and average pooling methods.
Further, in the dvCRNN network constructed in the step (3), after the training set signal is sent into the network, the training set signal is successively subjected to convolution layer, batch processing normalization layer and Relu layer, then the details extracted by the dual-view mechanism and the overall characteristics are fused together, the time sequence characteristics of the signal are extracted by using a gating cycle unit (GRU), and finally the signal is classified by using a full connection layer.
Advantageous effects
The invention provides a peripheral artery disease diagnosis method based on deep learning. Firstly, acquiring tissue blood flow change data of gastrocnemius muscle parts of a peripheral artery disease patient and a healthy volunteer based on a diffusion correlation spectroscopy technology; then, training the acquired data based on the proposed dvCRNN network, and extracting key features to obtain a deep learning network model for diagnosing peripheral artery diseases; and finally, sending the test set data into the trained network model to realize the diagnosis of peripheral artery diseases. The method is simple to use, does not need manual intervention, saves time and labor, overcomes the defect that the peripheral artery disease is diagnosed by measuring ABI parameters in the traditional method, verifies the feasibility of tissue blood flow measurement for peripheral artery disease diagnosis, and provides a new technology and a new method for peripheral artery disease diagnosis.
Drawings
FIG. 1 is an overall flow diagram of the process of the present invention;
FIG. 2 is a schematic diagram of the tissue blood flow change acquisition stages of the present invention;
FIG. 3 is data of changes in blood flow in tissue of a patient with peripheral arterial disease and healthy volunteers according to an embodiment of the present invention;
fig. 4 is a diagram of the deep learning model architecture according to the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention are described in further detail below with reference to the accompanying drawings in conjunction with specific embodiments.
The invention provides a peripheral artery disease diagnosis method based on deep learning, and FIG. 1 shows an overall flow chart of the method provided by the invention, and specifically comprises the following steps:
step (1): and (6) data acquisition. The over-diffusion related spectrum blood flow detection system acquires tissue blood flow change data of gastrocnemius muscle parts of a tested patient with peripheral artery disease and a healthy volunteer. During measurement, the testees are in a lying state, the probe is placed at the gastrocnemius part of each testee, and the cuff is placed at the root of the thigh of the lower limb measured by the testee and used for blocking blood flow. Fig. 2 shows a schematic diagram of each stage of data acquisition, and it can be seen that the data acquisition is mainly divided into three stages, wherein the stationary stage is 1min, the cuff occlusion stage is 5min, and the recovery stage is 10min. And in the stationary stage, the tested person keeps a relaxed state and lies down to measure the tissue blood flow. In the artery occlusion stage, the cuff pressure is instantaneously increased to be higher than the systolic pressure of the tested person and kept for 5min. The recovery phase is to release the cuff pressure to 0mmHg instantaneously and measure the blood flow change 10min after the release to ensure the blood flow to return to a steady state. The whole measuring process is 16min in total, and the number of sampling points is 150. The blood flow change data of the peripheral artery disease patients and healthy volunteers obtained by actual collection are shown in fig. 3, and can be obviously seen from the figure: in the stationary phase, the blood flow remains in a steady state; after the cuff pressurizes the occlusion blood flow, the blood flow value drops rapidly and remains low; after the cuff pressure is released, namely in a recovery stage, the blood flow value slowly falls back to the initial stable state after rising to a certain height. As can also be seen in the figure, there is a clear difference between the blood flow changes in patients with peripheral arterial disease and healthy volunteers: in the artery occlusion stage, the blood flow of the patient with peripheral artery disease is reduced to a lower degree than that of a healthy volunteer, and the blood flow value is higher than that of the healthy volunteer; in the recovery phase, the blood flow of patients with peripheral artery disease rises slowly and the peak blood flow value is significantly lower than that of healthy volunteers. It can be seen that tissue blood flow change detection based on the diffusion-related spectrum can clearly distinguish peripheral artery disease patients from healthy volunteers.
Step (2): and (4) preprocessing data. And (2) for the tissue blood flow change data obtained in the step (1), after normalization preprocessing, randomly dividing the tissue blood flow change data of the tested person into a training set and a testing set. And then, performing data enhancement on the training set samples by a data enhancement method, wherein the training set samples after the data enhancement are used as data input of the deep learning network. The collected tissue blood flow data of the patients with peripheral artery disease totally totaled 166 samples, and the data of healthy volunteers totally totaled 94 samples. Randomly selecting 130 samples from 166 peripheral artery disease patient tissue blood flow samples as training set samples, and taking the rest 36 samples as prediction set samples; similarly, 70 of 94 healthy volunteer data tissue blood flow samples were randomly selected as training set samples, and the remaining 24 were selected as prediction set samples. For 130 patients with peripheral artery disease except the training set and 70 healthy volunteers, the tissue blood flow data samples are respectively subjected to down-sampling to increase the sample signal to three sample signals, and then the up-sampling is used to restore the increased signal to the original length, namely 150 sample points. Therefore, the 130 patients with peripheral arterial disease and 70 healthy volunteers in the training set have their tissue blood flow data samples enhanced to 390 and 210 respectively.
And (3): and (5) constructing a deep learning network. Aiming at the characteristics of the collected tissue blood flow change data, the invention provides a dual-view convolution cyclic neural network (dvCRNN), which introduces a dual-view mechanism on the basis of a Convolution Neural Network (CNN) and is used for fully providing characteristic information in a sample. The double-view mechanism extracts and learns the characteristics of the blood flow signal from a detail layer and an integral layer, and obtains the details and the integral characteristics in the signal by respectively adopting a maximum pooling method and an average pooling method. After being sent to a network, the training set signals sequentially go through a convolutional layer, a normalization layer and a Relu layer, then details extracted by a double-view mechanism and overall characteristics are fused together, a gate control cycle unit (GRU) is used for extracting time sequence characteristics of the signals, and finally a full connection layer is used for classifying the signals. Fig. 4 shows an architecture diagram of a dual-view convolutional recurrent neural network (dvCRNN) proposed by the present invention.
And (4): and (5) network training and diagnosis. Inputting the enhanced sample data of the training set in the step (2) into the dvCRNN network constructed in the step (3) for training, and storing various parameters of the network after the training is finished to obtain a peripheral artery disease diagnosis model. And then, inputting the preprocessed sample data of the test set into a peripheral artery disease diagnosis model for diagnosis to obtain a diagnosis result.
Finally, it should be noted that although the present invention has been described with reference to the preferred embodiments, it should be understood by those skilled in the art that the above-mentioned preferred embodiments are merely illustrative of the present invention and are not intended to limit the scope of the present invention, and any modifications, equivalent substitutions, improvements and the like made within the spirit and scope of the present invention should be included in the scope of the present invention.
Claims (7)
1. A peripheral artery feature extraction method based on deep learning is characterized by comprising the following steps:
step (1): collecting data; collecting tissue blood flow change data of gastrocnemius muscle parts of a peripheral artery disease patient and a healthy volunteer through a diffusion related spectrum blood flow detection system; during measurement, the testees are in a lying state, the probe is placed at the gastrocnemius part of each tester, and the cuff is placed at the root of the thigh of the lower limb measured by the tester and used for blocking blood flow;
step (2): preprocessing data; after normalization preprocessing is carried out on the tissue blood flow change data obtained in the step (1), randomly dividing the tissue blood flow change data of a tested person into a training set and a testing set; then, performing data enhancement on the training set sample by a data enhancement method, wherein the training set sample after data enhancement is used as data input of a deep learning network;
and (3): constructing a deep learning network; aiming at the characteristics of collected tissue blood flow change data, the application provides a dual-view convolution cyclic neural network dual-view CRNN, wherein a dual-view mechanism is introduced into the network on the basis of a Convolution Neural Network (CNN) and is used for fully providing characteristic information in a sample; firstly, respectively inputting a light intensity autocorrelation function g2 (tau) into two parallel branches, wherein one branch is sequentially convolution and maximum pooling, and the other branch is sequentially convolution and average pooling, wherein the maximum pooling is used for extracting and learning detailed characteristics of blood flow signals, the average pooling is used for extracting and learning overall characteristics of blood flow, then the two branches are spliced, the spliced data is leveled and input into an RNN network gate control cycle unit (GRU), and the time sequence characteristics of the signals are extracted;
and (4): network training and feature extraction; inputting the enhanced training set sample data in the step (2) into the dvCRNN network constructed in the step (3) for training, and inputting the test set sample data into a trained network model for feature extraction after training is completed.
2. The method for extracting peripheral artery features based on deep learning as claimed in claim 1, wherein the data collection in step 1 is divided into three stages, wherein the stationary stage is 1min, the cuff occlusion stage is 5min, and the recovery stage is 10min.
3. The method for extracting peripheral artery features based on deep learning as claimed in claim 2, wherein during the data acquisition process, the measurement of tissue blood flow is performed in a steady stage, i.e. the tested person is kept in a relaxed lying state.
4. The method for extracting peripheral artery features based on deep learning as claimed in claim 2, wherein in the process of data acquisition, the artery occlusion stage is to increase the cuff pressure to be greater than the systolic pressure of the tested person instantaneously and keep it for 5min.
5. The method as claimed in claim 2, wherein during the data collection process, the cuff pressure is released to 0mmHg instantaneously, and the blood flow change is measured 10min after the cuff pressure is released to ensure that the blood flow returns to a steady state.
6. The method as claimed in claim 1, wherein the data enhancement method is to increase the number of sample signals to three by using a down-sampling method and then restore the increased sample signals to the original length by using an up-sampling method.
7. The method according to claim 1, wherein the data enhancement method is only applied to the training set samples and does not act on the test set samples.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011456466.4A CN112587119B (en) | 2020-12-11 | 2020-12-11 | Peripheral artery feature extraction method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011456466.4A CN112587119B (en) | 2020-12-11 | 2020-12-11 | Peripheral artery feature extraction method based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112587119A CN112587119A (en) | 2021-04-02 |
| CN112587119B true CN112587119B (en) | 2023-03-31 |
Family
ID=75192952
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011456466.4A Active CN112587119B (en) | 2020-12-11 | 2020-12-11 | Peripheral artery feature extraction method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112587119B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113116692B (en) * | 2021-04-19 | 2022-06-07 | 南京智能高端装备产业研究院有限公司 | Self-adaptive intermittent inflation pressurization system based on interface pressure |
| CN114081483A (en) * | 2021-12-28 | 2022-02-25 | 北京工业大学 | Tissue blood flow and blood oxygen saturation measuring method based on deep learning diffusion correlation spectrum |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108670240A (en) * | 2018-06-15 | 2018-10-19 | 中国工程物理研究院流体物理研究所 | The device and method of measurement biological tissue blood volume, blood oxygen, blood flow and oxygen metabolism |
| CN109316177A (en) * | 2018-11-02 | 2019-02-12 | 杜玲娟 | A kind of assessment device and method for peripheral arterial disease hemadostewnosis degree |
| CN110573066A (en) * | 2017-03-02 | 2019-12-13 | 光谱Md公司 | Machine learning systems and techniques for multi-spectral amputation site analysis |
| CN111358473A (en) * | 2020-03-17 | 2020-07-03 | 北京工业大学 | Tissue blood flow blood oxygen imaging device and method based on near infrared spectrum |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3622882A1 (en) * | 2018-09-14 | 2020-03-18 | Fundació Institut de Ciències Fotòniques | System and computer-implemented method for detecting and categorizing pathologies through an analysis of pulsatile blood flow |
-
2020
- 2020-12-11 CN CN202011456466.4A patent/CN112587119B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110573066A (en) * | 2017-03-02 | 2019-12-13 | 光谱Md公司 | Machine learning systems and techniques for multi-spectral amputation site analysis |
| CN108670240A (en) * | 2018-06-15 | 2018-10-19 | 中国工程物理研究院流体物理研究所 | The device and method of measurement biological tissue blood volume, blood oxygen, blood flow and oxygen metabolism |
| CN109316177A (en) * | 2018-11-02 | 2019-02-12 | 杜玲娟 | A kind of assessment device and method for peripheral arterial disease hemadostewnosis degree |
| CN111358473A (en) * | 2020-03-17 | 2020-07-03 | 北京工业大学 | Tissue blood flow blood oxygen imaging device and method based on near infrared spectrum |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112587119A (en) | 2021-04-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | The obf database: A large face video database for remote physiological signal measurement and atrial fibrillation detection | |
| Allen et al. | Development of a neural network screening aid for diagnosing lower limb peripheral vascular disease from photoelectric plethysmography pulse waveforms | |
| CN112587119B (en) | Peripheral artery feature extraction method based on deep learning | |
| Lin | Assessment of bilateral photoplethysmography for lower limb peripheral vascular occlusive disease using color relation analysis classifier | |
| Srinivasan et al. | Deep Learning based non-invasive diabetes predictor using Photoplethysmography signals | |
| Li et al. | Hybrid D1DCnet using forehead iPPG for continuous and noncontact blood pressure measurement | |
| Allen et al. | Prospective assessment of an artificial neural network for the detection of peripheral vascular disease from lower limb pulse waveforms | |
| CN113040738B (en) | Blood pressure detecting device | |
| Yen et al. | Applying a deep learning network in continuous physiological parameter estimation based on photoplethysmography sensor signals | |
| CN112587118B (en) | Diffusion-related spectrum blood flow quantification method based on deep learning | |
| CN104027097A (en) | Vascular function noninvasive detecting method and device | |
| JP4355386B2 (en) | Biological homeostasis maintenance evaluation device | |
| CN117017310A (en) | Acoustic-electric dual-mode congenital heart disease prediction device based on knowledge distillation | |
| Charlton et al. | Vascular biology and microcirculation: Assessing hemodynamics from the photoplethysmogram to gain insights into vascular age: A review from VascAgeNet | |
| Liu et al. | Convolutional neural network-based detection of deep vein thrombosis in a low limb with light reflection rheography | |
| Wu et al. | Noninvasive power spectrum analysis of radial pressure waveform for assessment of vascular system | |
| Sharpley et al. | Differences between ECG and pulse when measuring heart rate and reactivity under two physical and two psychological stressors | |
| Guevara et al. | Algorithm construction methodology for diagnostic classification of near-infrared spectroscopy data | |
| JP4801473B2 (en) | Exercise tolerance evaluation device | |
| JP6356246B2 (en) | Biological tissue diagnostic apparatus and method of operating biological tissue diagnostic apparatus | |
| Oliveira et al. | Machine learning-based diabetes detection using photoplethysmography signal features | |
| CN113712524B (en) | Data processing device, system and kit for auxiliary assessment of cardiovascular risk | |
| Kochev et al. | A Novel Approach for Modelling the Relationship between Blood Pressure and ECG by using Time-series Feature Extraction | |
| Nuryani et al. | Advancing Non-Cuff Hypertension Detection: Leveraging 1D Convolutional Neural Network and Time Domain Physiological Signals. | |
| Masdoua et al. | Blood Pressure Assessment from Contact Photoplethysmographic Signals Using a Combination of Deep Convolutional and Recurrent Neural Networks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |