Bridge vessel blood flow measurement waveform analysis method and device based on time-frequency characteristics

Abstract

The invention discloses a bridge vessel blood flow measurement waveform analysis method and device based on time-frequency characteristics, and relates to the technical field of machine learning. Comprising the following steps: collecting an instant blood flow measurement waveform of a bridge blood vessel, and extracting a plurality of time domain features from the instant blood flow measurement waveform; transforming an instant blood flow measurement waveform from a time domain to a frequency domain through discrete Fourier transform, discrete cosine transform and discrete wavelet transform; extracting a plurality of frequency domain features from the frequency domain; constructing a data set based on the time domain features and the frequency domain features, constructing a deep neural network, and training the deep neural network by using the data set to obtain a bridge vessel patency analysis model; and analyzing the immediate patency of the bridge blood vessel and the postoperative middle and long-term patency by using the bridge blood vessel patency analysis model.

Description

Technical Field

[0001] The present invention relates to the technical field of machine learning, and particularly relates to a method and device for analyzing the blood flow measurement waveform of a bypass vessel based on time-frequency characteristics. Background Art

[0002] Coronary artery bypass grafting (CABG) is a common surgical method for treating coronary heart disease. Doctors intercept a blood vessel from other parts of the patient's own body as a bypass vessel, connect it to both ends of the coronary artery stenosis site, and make the blood flow through the bypass vessel to perfuse the distal end of the coronary artery, so as to achieve the purpose of treating myocardial ischemia. According to the different bypass vessel materials selected, it can be divided into internal mammary artery bypass vessel (LIMA) and saphenous vein bypass vessel (SVG). An important problem in bypass surgery is the patency of the bypass vessel, that is, the bypass vessel may become occluded some time after the operation, which will again threaten the life and health of the patient. The patency of the bypass vessel can be divided into immediate patency and medium- and long-term patency after the operation. The immediate patency problem refers to the patency problem that occurs within a very short time after the operation (within one week after the operation), mostly caused by the surgical operation method; the medium- and long-term patency refers to the patency problem that occurs more than one year after the operation, and the patient's own physical state and hemodynamic environment account for a large reason. The accurate prediction of these two kinds of patency is an urgent problem to be solved in clinical practice.

[0003] Currently, a method called trans-thoracic flow measurement (TTFM) is commonly used in clinical practice to evaluate the blood flow status of the bypass vessel after bypass surgery, and is also often used to predict the patency of the bypass vessel. The TTFM method is to clamp an ultrasonic probe on the bypass vessel before closing the chest after coronary artery bypass surgery to measure the flow waveform of the bypass vessel, which is the TTFM waveform. From the TTFM waveform, clinical features such as mean flow, pulsatility index (PI), and diastolic flow ratio (DF) can be read out, and these features are often referred to by doctors at present to evaluate the flow status of the bypass vessel. The process of measuring the waveform is as Figure 1 shown.

[0004] Currently, in the analysis of bypass vessel patency, the mainstream methods tend to use one or several clinical features, establish a discriminant formula or a logistic regression model, and set a specific threshold to classify and judge the patency. However, the accuracy of the analysis results of these methods needs to be improved. Summary of the Invention

[0005] Based on this, in order to solve the technical problems in the prior art, the present invention provides a method and device for analyzing the blood flow measurement waveform of a bypass vessel based on time-frequency characteristics.

[0006] The present invention provides a method for analyzing the blood flow measurement waveform of a bypass vessel based on time-frequency characteristics, including:

[0007] Collect the instantaneous blood flow measurement waveform of the bypass vessel and the corresponding bypass vessel patency label;

[0008] Respectively transform the instantaneous blood flow measurement waveform from the time domain to the first frequency domain, the second frequency domain, and the third frequency domain through discrete Fourier transform, discrete cosine transform, and discrete wavelet transform; and respectively extract the first frequency domain characteristic parameter vector, the second frequency domain characteristic parameter vector, and the third frequency domain characteristic parameter vector reflecting the global periodicity, low-frequency domain characteristics, and instantaneous frequency domain characteristics of the instantaneous blood flow measurement waveform from the first frequency domain, the second frequency domain, and the third frequency domain;

[0009] Extract the time domain characteristic parameter vector from the instantaneous blood flow measurement waveform through time-frequency analysis methods;

[0010] Construct a data set based on the first frequency domain characteristic parameter vector, the second frequency domain characteristic parameter vector, the third frequency domain characteristic parameter vector, the time domain characteristic parameter vector, and the corresponding bypass vessel patency label, construct a deep neural network, use the data set to train the deep neural network to obtain a bypass vessel blood flow measurement waveform analysis model; use the bypass vessel blood flow measurement waveform analysis model to analyze the bypass vessel blood flow measurement waveform.

[0011] Further, the parameters included in the first frequency domain characteristic parameter vector, the second frequency domain characteristic parameter vector, and the third frequency domain characteristic parameter vector are the same, and all include: amplitude average value, center frequency, mean square frequency, variance frequency, frequency variance, average frequency, median frequency, total power, average power, frequency corresponding to the maximum power value, and low-frequency / high-frequency power.

[0012] Further, the time domain characteristic parameter vector includes average flow rate, pulsatility index, diastolic flow rate ratio, and negative flow ratio.

[0013] Further, the construction of the data set based on the first frequency domain characteristic parameter vector, the second frequency domain characteristic parameter vector, the third frequency domain characteristic parameter vector, the time domain characteristic parameter vector, and the corresponding bypass vessel patency label specifically includes:

[0014] Use the 4 time domain characteristic parameters in the time domain characteristic parameter vector, the 11 frequency domain characteristic parameters in the first frequency domain characteristic parameter vector, the 11 frequency domain characteristic parameters in the second frequency domain characteristic parameter vector, and the 11 frequency domain characteristic parameters in the third frequency domain characteristic parameter vector, a total of 37 characteristic parameters as the input of the deep neural network, and the bypass vessel patency label corresponding to the 37 characteristic parameters as the label of the deep neural network to construct a data set.

[0015] Further, the bypass vessel patency label includes the bypass vessel instantaneous patency label and the bypass vessel medium- and long-term patency label.

[0016] Further, the training of the deep neural network using the data set specifically includes:

[0017] Set the loss function used in the training process of the deep neural network as the cross - entropy loss function, and the optimization method as the stochastic gradient descent method;

[0018] Set the hyperparameters of the deep neural network and use grid search to tune the hyperparameters;

[0019] During the training process, use k - fold cross - validation to tune the weight parameters of the deep neural network.

[0020] The present invention provides a device for analyzing the blood flow measurement waveform of a bypass vessel based on time - frequency characteristics, including:

[0021] A data acquisition module, configured to acquire the instant blood flow measurement waveform of the bypass vessel and the corresponding bypass vessel patency label;

[0022] A frequency - domain feature extraction module, configured to transform the instant blood flow measurement waveform from the time domain to the first frequency domain, the second frequency domain, and the third frequency domain through discrete Fourier transform, discrete cosine transform, and discrete wavelet transform respectively; and extract the first frequency - domain feature parameter vector, the second frequency - domain feature parameter vector, and the third frequency - domain feature parameter vector reflecting the global periodicity, low - frequency domain characteristics, and instantaneous frequency domain characteristics of the instant blood flow measurement waveform from the first frequency domain, the second frequency domain, and the third frequency domain respectively;

[0023] A time - domain feature extraction module, configured to extract a time - domain feature parameter vector from the instant blood flow measurement waveform through a time - frequency analysis method;

[0024] A data analysis module, configured to construct a data set based on the first frequency - domain feature parameter vector, the second frequency - domain feature parameter vector, the third frequency - domain feature parameter vector, the time - domain feature parameter vector, and the corresponding bypass vessel patency label, construct a deep neural network, train the deep neural network using the data set to obtain a bypass vessel blood flow measurement waveform analysis model; and analyze the bypass vessel blood flow measurement waveform using the bypass vessel blood flow measurement waveform analysis model.

[0025] The above - mentioned at least one technical solution adopted by the present invention can achieve the following beneficial effects:

[0026] In the method for analyzing the blood flow measurement waveform of the bypass vessel based on time-frequency characteristics provided by the present invention, a multi-scale frequency-domain analysis method is used to map the time domain to the frequency domain, and the frequency-domain characteristics and time-domain characteristics are combined and then input into a deep neural network for analysis, which can more accurately capture and analyze the characteristics of the blood flow volume waveform and obtain more accurate analysis results. Specifically: The discrete Fourier transform focuses on global frequency-domain analysis and can provide the overall frequency information of the blood flow waveform; the discrete cosine transform compresses the blood flow waveform and particularly focuses on the analysis of low-frequency signals, which is suitable for extracting the key low-frequency characteristics in the blood flow waveform; the discrete wavelet transform focuses on the multi-resolution local frequency-domain analysis of different scales and can effectively capture the details and instantaneous changes in the blood flow waveform; these three methods analyze the blood flow waveform from different angles and jointly constitute a complementary analysis framework, which can help to more accurately capture and analyze the characteristics of the blood flow volume waveform and obtain more accurate analysis results. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention. In the drawings:

[0028] Figure 1 is a schematic diagram of the measurement waveform provided by the present invention; Figure 1 in (a) is a schematic diagram of the main unit of TTFM; Figure 1 in (b) is a schematic diagram of the process of measuring the blood flow waveform of the graft using a TTFM probe; Figure 1 in (c) is the measured TTFM waveform diagram;

[0029] Figure 2 is a schematic diagram of the flow chart of the method for analyzing the blood flow measurement waveform of the bypass vessel based on time-frequency characteristics provided by the present invention;

[0030] Figure 3 is a schematic diagram of the time-frequency conversion of the TTFM waveform diagram provided by the present invention, Figure 3 in (a) is the TTFM waveform diagram; Figure 3 in (b) is based on Figure 3 in (a) after the fast Fourier transform of the spectrogram;

[0031] Figure 4 is a spectrogram obtained based on the discrete Fourier transform provided by the present invention;

[0032] Figure 5 is a spectrogram obtained based on the discrete cosine transform provided by the present invention

[0033] Figure 6 is a spectrogram obtained based on the discrete wavelet transform provided by the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0034] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only a part rather than all of the embodiments of the present invention. Based on the embodiments in the specification, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0035] In current solutions, most doctors directly use the clinical features of the TTFM waveform for self-judgment, and there is currently no unified judgment standard. Common clinical features include mean flow, PI (calculation method: (maximum flow - minimum flow) / mean flow), DF (calculation method: diastolic flow / total flow), and proportion of reverse flow (reverse flow / total flow). Generally, it is considered that when the mean flow > 15 ml / min, PI < 5, DF > 50%, and there is no reverse flow, the blood flow status of the bypass vessel is considered good; otherwise, it is considered that there is a problem with the blood flow status of the bypass vessel.

[0036] Some scholars have predicted the immediate and medium- to long-term patency of bypass vessels based on the above-mentioned clinical features of the TTFM waveform. However, the methods used are either simply finding a threshold (cut-off value) for binary classification or combining several features to establish a discriminant equation. The prediction effects are not very good. Some scholars have also proposed to incorporate the frequency domain features of the TTFM waveform into the consideration of bypass vessel patency judgment. The scholar's research found that after performing a fast Fourier transform on the TTFM waveform, a spectrogram can be obtained, as Figure 3 shown, and its fundamental wave (F0), first harmonic (H1), second harmonic (H2), etc. can be extracted. Among them, using the ratio of the fundamental wave to the first harmonic (F0 / H1) as a feature to judge bypass vessel patency has a better effect than clinical features.

[0037] Based on the above research status, the disadvantages of the existing technical solutions are summarized as follows:

[0038] 1. Most of the current methods for predicting bypass vessel patency select one or several clinical features and choose a threshold to judge patency or non-patency. A few combine several clinical features to establish a discriminant formula or logistic regression formula for binary classification. The prediction effects of these methods are not very satisfactory. 2. Some scholars have proposed performing a fast Fourier transform on the TTFM, and then using the ratio of the fundamental wave to the harmonic as a judgment feature, but it also uses the method of selecting a threshold for simple judgment, and its prediction effect is not very good.

[0039] In this context, the present invention provides a method for analyzing the blood flow measurement waveform of bypass vessels based on time-frequency characteristics. By using various time-frequency transformation methods (discrete Fourier transform, discrete cosine transform, discrete wavelet transform), multiple frequency-domain characteristics (amplitude average value, center frequency, mean square frequency, variance frequency, frequency variance, average frequency, median frequency, total power, average power, frequency corresponding to the maximum power value, low-frequency / high-frequency power ratio) are extracted and fused with the clinical characteristics in the time domain to jointly serve as the characteristics of the prediction model. At the same time, a deep neural network is selected as the method for the prediction model to construct a bypass vessel patency prediction model, realizing the patency prediction (immediate patency, medium- and long-term patency) of different types of bypass vessels (LIMA, SVG) at different postoperative periods.

[0040] Example 1

[0041] Figure 2 The flowchart of the method for analyzing the blood flow measurement waveform of bypass vessels based on time-frequency characteristics in this embodiment is shown. The following specifically describes this method in detail, including the following steps: Figure 2 This method is specifically described in detail as follows, including the following steps:

[0042] S1: Collect the immediate blood flow measurement waveform of the bypass vessel and the corresponding bypass vessel patency label.

[0043] Before closing the chest after coronary artery bypass grafting, use an ultrasonic probe to clamp on the bypass vessel to collect the immediate blood flow measurement waveform of the bypass vessel, as shown in (c) of Figure 1 or (a) of Figure 3 as shown.

[0044] S2: Respectively transform the immediate blood flow measurement waveform from the time domain to the first frequency domain, the second frequency domain, and the third frequency domain through discrete Fourier transform, discrete cosine transform, and discrete wavelet transform; and respectively extract the first frequency domain characteristic parameter vector, the second frequency domain characteristic parameter vector, and the third frequency domain characteristic parameter vector reflecting the global periodicity, low-frequency domain characteristics, and instantaneous frequency domain characteristics of the immediate blood flow measurement waveform from the first frequency domain, the second frequency domain, and the third frequency domain.

[0045] S201: Perform time-frequency transformation on the TTFM waveform.

[0046] Perform time-frequency transformation on the TTFM waveform. The time-frequency transformation methods used are discrete Fourier transform, discrete cosine transform, and discrete wavelet transform, as shown in Figures 4 to 6 as shown.

[0047] Frequency domain representation can usually effectively reduce redundant information and improve the representation efficiency of signals. Discrete Fourier Transform (DFT), Discrete Cosine Transform (DCT), and Discrete Wavelet Transform (DWT) are three time-frequency transformation methods that can convert signals from the time domain to the frequency domain to analyze the characteristics of signals at different frequency components. Through frequency domain analysis, periodic characteristics, frequency distribution, and energy distribution in signals can be identified more clearly. DFT focuses on global frequency domain analysis and can provide the overall frequency information of signals. DCT further compresses the signals, especially focusing on the analysis of low-frequency signals, and is suitable for extracting key low-frequency characteristics in signals. DWT focuses on multi-resolution local frequency domain analysis at different scales and can effectively capture details and instantaneous changes in signals. These three methods analyze signals from different perspectives and can complement each other. The combined time-frequency transformation method will have more advantages than a single time-frequency analysis method, and the combination of this information can comprehensively improve the detection performance.

[0048] Blood flow waveforms usually contain periodic changes, instantaneous fluctuations, and complex frequency components. For periodic characteristics, the frequency components in blood flow waveforms may change with different physiological conditions. Through DFT analysis, the main frequency components in the waveform can be identified to help diagnose cardiovascular diseases or hemodynamic changes. The low-frequency components in blood flow waveforms usually contain most of the main characteristics of blood flow. DCT can effectively extract these key low-frequency information while reducing the interference of high-frequency noise. For the possible instantaneous changes in blood flow waveforms within a short period, which usually manifest as sudden changes or fluctuations in the signal, DWT can analyze these changes at multiple resolution levels by decomposing the signal into sub-signals of different frequency bands, ensuring effective capture of both fast and slow-changing signals. This multi-scale analysis can reveal subtle changes in the blood flow waveform. Combining the advantages of these three time-frequency transformation methods can help capture and analyze the characteristics of blood flow waveforms more accurately, thereby improving the accuracy of monitoring and diagnosis.

[0049] Through the Discrete Fourier Transform, a discrete signal of a finite-length sequence in the time domain can be converted to the frequency domain, thus better analyzing and understanding various characteristics and change laws of the signal. Specifically, the Discrete Fourier Transform can convert a complex discrete signal of a finite-length sequence into a linear combination of simple sine and cosine signals, and convert it into the corresponding amplitude and phase at different frequencies, enabling the acquisition of the frequency components contained in the signal as a whole, separating the spectral information of each dimension, and analyzing the spectral structure and change laws of the signal, so as to understand the signal more comprehensively.

[0050] The discrete cosine transform is a transform similar to the discrete Fourier transform. It is derived based on the discrete Fourier transform and is essentially also a discrete Fourier transform. It also uses a triangular wave function for the time-frequency transformation of signals and is a special form of the discrete Fourier transform. The difference is that the discrete cosine transform is performed on real even functions and uses the cosine function instead of the sine function for the spectral analysis of signals.

[0051] The discrete wavelet transform is a signal decomposition method based on wavelet analysis that can discretize the scale and translation of the basic wavelet. The wavelet transform decomposes the signal into wavelet components of different frequencies and extracts the local time-frequency information of the signal. It can comprehensively analyze the signal at different scales and frequencies, being able to examine both the frequency-domain characteristics of the local time-frequency process and the time-frequency characteristics of the local frequency-domain process.

[0052] S202: Extract frequency-domain features from the frequency domain.

[0053] After obtaining the frequency information of the TTFM waveform by using different time-frequency transformation methods, in order to analyze the specific meaning of the frequency information, we use the following several frequency-domain feature variables for disease prediction: amplitude average value, center frequency, mean square frequency, variance frequency, frequency variance, average frequency, median frequency, total power, average power, frequency corresponding to the maximum power value, low-frequency / high-frequency power ratio.

[0054] The amplitude average value of the frequency is the average value of the amplitudes of the signal within a certain frequency range over a period of time; the center frequency refers to the center of the signal spectrum; the mean square frequency refers to the weighted average of the product of the amplitudes and frequencies of each frequency component in the signal spectrum; the variance frequency refers to the variance of the frequency distribution; the frequency variance is used to measure the degree of dispersion of the frequency distribution and is the weighted average of the squares of the differences between each frequency and the center frequency; the average frequency refers to the central position of the frequency distribution; the median frequency refers to the middle value of the frequency distribution in the signal spectrum; the total power refers to the total energy of the signal within all frequency ranges; the average power refers to the average energy of the signal within all frequency ranges; the frequency corresponding to the maximum power value refers to the frequency with the maximum power value in the signal spectrum; the low-frequency / high-frequency power ratio refers to the ratio of the signal power in the low-frequency and high-frequency ranges and is used to understand the energy distribution of the signal in different frequency ranges.

[0055] S3: Extract the time-domain feature parameter vector from the instantaneous blood flow measurement waveform through time-frequency analysis methods.

[0056] Common clinical features include mean flow, PI (calculated as: (maximum flow - minimum flow) / mean flow), DF (calculated as: diastolic flow / total flow), and the proportion of negative flow (negative flow / total flow). Generally, if the mean flow > 15 ml / min, and PI < 5, and DF > 50%, and there is no negative flow, it is considered that the blood flow status of the bypass vessel is good; otherwise, it is considered that there is a problem with the blood flow status of the bypass vessel.

[0057] S4: Construct a data set based on the first frequency-domain feature parameter vector, the second frequency-domain feature parameter vector, the third frequency-domain feature parameter vector, the time-domain feature parameter vector, and the corresponding bypass vessel patency label, construct a deep neural network, use the data set to train the deep neural network to obtain a bypass vessel blood flow measurement waveform analysis model; use the bypass vessel blood flow measurement waveform analysis model to analyze the bypass vessel blood flow measurement waveform.

[0058] S401: Construct a data set based on time-domain features and frequency-domain features.

[0059] The characteristic information used in the present invention includes 4 clinical basic characteristics: mean flow, PI, DF, and the proportion of negative flow. The frequency characteristics include 11 characteristics such as the average amplitude, center frequency, mean square frequency, variance frequency, average frequency, frequency variance, median frequency, total power, average power, the frequency corresponding to the maximum power value, and the low-frequency / high-frequency power ratio obtained by Fourier transform, cosine transform, and wavelet transform respectively, totaling 33 characteristics. The combined 37 characteristics of the two are used as the characteristics for constructing the prediction model.

[0060] S302: Use the data set to train the deep neural network.

[0061] The present invention uses a deep neural network as an artificial intelligence method for predicting the patency of bypass vessels. A deep neural network is a multi-layer neural network, and its basic unit is a neuron. A neuron receives multiple inputs, performs some mathematical operations, and then generates an output of the unit. In a deep neural network, each neuron is connected to all neurons in the previous layer and all neurons in the next layer, and is trained through the backpropagation algorithm to finally achieve the classification or regression of input data.

[0062] Use nn.Sequential in the PyTorch deep learning framework to build a multi-layer neural network. The number of layers affects the complexity of the model. More layers can improve the expressive ability of the model, but too many layers will cause problems such as gradient vanishing, gradient explosion, and overfitting. To solve the above problems, activation functions are added, and at the same time, non-linearity is introduced so that the neural network can learn more complex patterns. The batch size is the number of samples used for training in each iteration. The loss function measures the gap between the predicted value and the true value of the model. The optimizer is used to update the model parameters to minimize the loss function. The learning rate controls the step size of parameter update, and the number of iterations (Epoch) of the dataset affects the final training performance of the model. Usually, enough iterations are required to make the model converge.

[0063] In the present invention, a classification model is constructed using the Scikit-learn library of Python and the PyTorch deep learning framework, and the grid search and cross-validation methods are used to optimize the model parameters. Grid search is a common method for hyperparameter tuning of the model. Based on exhaustive search, by specifying a list of candidate values for the parameters, all possible parameter combinations are evaluated and compared to select the best parameter combination. k-fold cross-validation is a common technique for evaluating the performance of the model. It divides the dataset into a training set and a validation set and repeats this division k times. In each division, the training set is used to train the model, and the validation set is used to evaluate the performance of the model. Finally, the average performance of the model is evaluated based on the accuracy results of all cross-validations. In the present invention, according to the number of data samples, 5-fold cross-validation is used to evaluate the model performance. By comparing the accuracy results of cross-validation under all parameter combinations, the parameter combination corresponding to the maximum average accuracy is the optimal parameter.

[0064] In the present invention, the learning rate of the deep neural network model is set to 0.001, and the epoch is set to 300. Since the scale of the dataset itself is small, better performance can be achieved when the batch size is set to 8. The Relu function is used as the activation function, the cross entropy is used as the loss function, and the stochastic gradient descent method is used as the optimizer to optimize the loss function. The random seed is set to 1.

[0065] S303: Tuning of the deep neural network.

[0066] The trained predictor is evaluated using five metrics: precision, recall, F1 score, accuracy, and AUC value. Precision represents the proportion of samples predicted as positive by the model that are truly positive. Recall represents the proportion of positive samples correctly predicted by the model among all true positive samples. The F1 score is a metric that comprehensively considers precision and recall and is used to evaluate the overall performance of a classification model. The F1 score is the harmonic mean of precision and recall, that is, the combined performance of the model's precision and recall. Accuracy represents the proportion of samples correctly predicted by the model among the total number of samples. The AUC value is the area under the ROC curve. The AUC value is usually used to judge the quality of a classifier, and its value ranges from 0 to 1. The larger the AUC value, the better the classification performance of the model.

[0067] The effectiveness of the present invention has been verified through experiments. We collected the TTFM waveforms of 796 arterial grafts (LIMA) and 995 venous grafts (SVG), as well as the coronary CTA results one week after surgery, to verify the prediction effect of the model on immediate patency. We collected the TTFM waveforms of 210 arterial grafts and 268 venous grafts, as well as the coronary CTA results one year after surgery, to verify the prediction effect of the model on mid- and long-term patency. The results are shown in Table 1. As can be seen from the table, our method has very good prediction effects on both arterial grafts and venous grafts, for both immediate patency and mid- and long-term patency.

[0068] Table 1 Model Prediction Results

[0069]

[0070] The above method for analyzing the blood flow measurement waveform of bypass vessels based on time-frequency features realizes the prediction of immediate patency and mid- and long-term patency of bypass vessels after coronary artery bypass grafting, and can achieve a very high accuracy. It can help doctors timely revise the surgical strategy and formulate a more scientific and reasonable postoperative review and rehabilitation plan for patients.

[0071] The above is the method for analyzing the blood flow measurement waveform of bypass vessels based on time-frequency features provided by one or more embodiments of the present invention. Based on the same idea, the present invention also provides a corresponding device for analyzing the blood flow measurement waveform of bypass vessels based on time-frequency features, including:

[0072] A data acquisition module for acquiring the immediate blood flow measurement waveform of the bypass vessel and the corresponding bypass vessel patency label.

[0073] A frequency-domain feature extraction module, which is used to transform the instantaneous blood flow measurement waveform from the time domain to the first frequency domain, the second frequency domain, and the third frequency domain respectively through discrete Fourier transform, discrete cosine transform, and discrete wavelet transform; and extract the first frequency-domain feature parameter vector, the second frequency-domain feature parameter vector, and the third frequency-domain feature parameter vector that reflect the global periodicity, low-frequency domain features, and instantaneous frequency domain features of the instantaneous blood flow measurement waveform from the first frequency domain, the second frequency domain, and the third frequency domain respectively.

[0074] A time-domain feature extraction module, which is used to extract a time-domain feature parameter vector from the instantaneous blood flow measurement waveform through a time-frequency analysis method.

[0075] A data analysis module, which is used to construct a data set based on the first frequency-domain feature parameter vector, the second frequency-domain feature parameter vector, the third frequency-domain feature parameter vector, the time-domain feature parameter vector, and the corresponding bypass vessel patency label, construct a deep neural network, train the deep neural network using the data set to obtain a bypass vessel blood flow measurement waveform analysis model; and analyze the bypass vessel blood flow measurement waveform using the bypass vessel blood flow measurement waveform analysis model.

[0076] For the specific limitations of the bypass vessel blood flow measurement waveform analysis device based on time-frequency features, reference can be made to the limitations of the bypass vessel blood flow measurement waveform analysis method based on time-frequency features in the above text, which will not be elaborated here. Each module in the above bypass vessel blood flow measurement waveform analysis device based on time-frequency features can be implemented in whole or in part through software, hardware, and their combination. The above modules can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so as to facilitate the processor to call and execute the operations corresponding to the above modules.

[0077] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided by the present invention can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0078] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded by the present invention.

Claims

1. A waveform analysis method for measuring blood flow in a bridge vessel based on time-frequency characteristics, characterized in that: include: Collecting the real-time blood flow measurement waveform of the graft vessel and the corresponding graft vessel patency label; The instant blood flow measurement waveform is transformed from the time domain to the first frequency domain, the second frequency domain and the third frequency domain by discrete Fourier transform, discrete cosine transform and discrete wavelet transform respectively; and the first frequency domain characteristic parameter vector, the second frequency domain characteristic parameter vector and the third frequency domain characteristic parameter vector reflecting the global periodicity, low-frequency frequency domain characteristics and instantaneous frequency domain characteristics of the instant blood flow measurement waveform are extracted from the first frequency domain, the second frequency domain and the third frequency domain respectively; The time-domain characteristic parameter vector is extracted from the real-time blood flow measurement waveform by using the time-frequency analysis method; A data set is constructed based on the first frequency domain characteristic parameter vector, the second frequency domain characteristic parameter vector, the third frequency domain characteristic parameter vector, the time domain characteristic parameter vector and the corresponding bridge vessel patency label, a deep neural network is constructed, the deep neural network is trained using the data set, and a bridge vessel blood flow measurement waveform analysis model is obtained; the bridge vessel blood flow measurement waveform is analyzed using the bridge vessel blood flow measurement waveform analysis model.

2. The method for analyzing the waveform of blood flow measurement of bridge vessels based on time-frequency characteristics according to claim 1, characterized in that: The parameters contained in the first frequency domain characteristic parameter vector, the second frequency domain characteristic parameter vector and the third frequency domain characteristic parameter vector are the same, including: amplitude average, centroid frequency, mean square frequency, variance frequency, frequency variance, average frequency, median frequency, total power, average power, frequency corresponding to the maximum power value and low frequency / high frequency power.

3. The method for analyzing the waveform of blood flow measurement of bridge vessels based on time-frequency characteristics according to claim 1, characterized in that: The time domain characteristic parameter vector includes average flow, pulsatility index, diastolic flow ratio and negative flow ratio.

4. The method for analyzing the waveform of blood flow measurement of bridge vessels based on time-frequency characteristics according to claim 1, characterized in that: The step of constructing a data set based on the first frequency domain feature parameter vector, the second frequency domain feature parameter vector, the third frequency domain feature parameter vector, the time domain feature parameter vector and the corresponding bridge vessel patency label specifically includes: A total of 37 feature parameters, including 4 time domain feature parameters in the time domain feature parameter vector, 11 frequency domain feature parameters in the first frequency domain feature parameter vector, 11 frequency domain feature parameters in the second frequency domain feature parameter vector, and 11 frequency domain feature parameters in the third frequency domain feature parameter vector, were used as the input of the deep neural network, and the bridge vessel patency labels corresponding to the 37 feature parameters were used as the labels of the deep neural network to construct a data set.

5. The method for analyzing the waveform of blood flow measurement of bridge vessels based on time-frequency characteristics according to claim 1, characterized in that: The graft vessel patency label includes a graft vessel immediate patency label and a graft vessel mid- to long-term patency label.

6. The method for analyzing the waveform of blood flow measurement of bridge vessels based on time-frequency characteristics according to claim 1, characterized in that: The use of the data set to train the deep neural network specifically includes: The loss function used in the deep neural network training process is set to the cross entropy loss function, and the optimization method is the stochastic gradient descent method; Set the hyperparameters of the deep neural network and tune them using grid search; During the training process, k-fold cross validation is used to tune the weight parameters of the deep neural network.

7. A waveform analysis device for measuring blood flow in a bridge vessel based on time-frequency characteristics, characterized in that: include: A data acquisition module, used to acquire the real-time blood flow measurement waveform of the graft vessel and the corresponding graft vessel patency label; A frequency domain feature extraction module is used to transform the instant blood flow measurement waveform from the time domain to the first frequency domain, the second frequency domain and the third frequency domain respectively through discrete Fourier transform, discrete cosine transform and discrete wavelet transform; and extract the first frequency domain feature parameter vector, the second frequency domain feature parameter vector and the third frequency domain feature parameter vector reflecting the global periodicity, low-frequency frequency domain characteristics and instantaneous frequency domain characteristics of the instant blood flow measurement waveform from the first frequency domain, the second frequency domain and the third frequency domain respectively; A time domain feature extraction module is used to extract a time domain feature parameter vector from the real-time blood flow measurement waveform by using a time-frequency analysis method; The data analysis module is used to construct a data set based on the first frequency domain characteristic parameter vector, the second frequency domain characteristic parameter vector, the third frequency domain characteristic parameter vector, the time domain characteristic parameter vector and the corresponding bridge vessel patency label, construct a deep neural network, use the data set to train the deep neural network, and obtain a bridge vessel blood flow measurement waveform analysis model; use the bridge vessel blood flow measurement waveform analysis model to analyze the bridge vessel blood flow measurement waveform.

Images