Chronic heart failure traditional Chinese medicine syndrome type identification method and device
By dividing heart rate variability data into time and frequency domains, a multimodal model was constructed to predict TCM syndrome types, overcoming the limitations of single-modal data and improving the prediction accuracy and robustness of TCM syndrome types in chronic heart failure.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-26
- Publication Date
- 2026-03-13
AI Technical Summary
Existing methods for predicting TCM syndrome types are usually limited to single-modal data processing, resulting in inaccurate TCM syndrome type predictions and a failure to fully integrate multi-source data.
By acquiring time-domain and frequency-domain data of heart rate variability, CNN and Transformer models are constructed and trained respectively. By combining LSTM and fully connected modules, the temporal and global dependencies of the data are captured, and TCM syndrome prediction is performed using weighted summation and maximum probability selection methods.
It improves the recognition accuracy of TCM syndrome texts, automatically corrects misjudgments in single models, and enhances the accuracy and robustness of predictions.
Smart Images
Figure CN121662340A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of biological data analysis technology, and in particular to a method and device for identifying TCM syndrome types in chronic heart failure. Background Technology
[0002] Chronic heart failure (CHF) is the end stage of various heart diseases. Traditional Chinese medicine (TCM) classifies CHF into four syndrome types: Qi deficiency and blood stasis syndrome, Qi and Yin deficiency syndrome, heart and kidney Yang deficiency syndrome, and Yang deficiency with water retention syndrome. If the TCM syndrome types of CHF can be predicted, the treatment effect of CHF can be greatly improved.
[0003] With the development of deep learning, deep learning models are increasingly being applied to the prediction of TCM syndrome types in chronic heart failure. Existing methods for predicting TCM syndrome types typically train deep learning models on single-modal data, allowing the trained models to predict syndrome types based on that single-modal data. However, TCM syndrome types are often related to multiple factors; for example, time-domain data and frequency data have different impacts on TCM syndrome types. Considering only time-domain or frequency data will lead to inaccurate predictions. Therefore, existing prediction methods, limited to single-modal data processing, fail to fully integrate multi-source data, resulting in insufficient prediction accuracy. Summary of the Invention
[0004] To address the above problems, this invention provides a method for identifying TCM syndrome types in chronic heart failure, comprising the following steps: Acquire a set of heart rate variability data and a set of TCM syndrome types for chronic heart failure. The heart rate variability data set includes a time domain data set and a frequency domain data set. Obtain the first mapping relationship between the time-domain data set and the TCM syndrome type text set of chronic heart failure, and use the first mapping relationship as a label to annotate the TCM syndrome type text set of chronic heart failure to obtain the first label chronic heart failure TCM syndrome type text set. Obtain the second mapping relationship between the frequency domain data set and the TCM syndrome type text set of chronic heart failure. Use the second mapping relationship as a label to annotate the TCM syndrome type text set of chronic heart failure, and obtain the second label TCM syndrome type text set of chronic heart failure. The first initial TCM syndrome identification model is iteratively trained using a time-domain dataset and a first-label TCM syndrome text set for chronic heart failure. After training, the first final TCM syndrome identification model is obtained. The second initial TCM syndrome identification model is iteratively trained using a frequency domain dataset and a second-label TCM syndrome text set for chronic heart failure. After training, the second final TCM syndrome identification model is obtained. A set of heart rate variability data to be tested is obtained, and the set of heart rate variability data to be tested is divided into a time domain data set and a frequency domain data set to be tested. The time domain data set is input into the first final TCM syndrome identification model for TCM syndrome prediction to obtain the first TCM syndrome probability set; the frequency domain data set is input into the second final TCM syndrome identification model for TCM syndrome prediction to obtain the second TCM syndrome probability set. The first set of TCM syndrome probabilities and the second set of TCM syndrome probabilities are weighted and summed, and the maximum probability is selected to obtain the final TCM syndrome probability. The final TCM syndrome text of chronic heart failure is obtained based on the final TCM syndrome probability.
[0005] Optional: The time-domain dataset consists of multiple time-domain data sets, and each time-domain data set includes one SDNN parameter and one rMSSD parameter. The frequency domain dataset consists of multiple frequency domain data sets. Each frequency domain data set includes an LF parameter, an HF parameter, and an LF / HF parameter. The collection of TCM syndrome texts for chronic heart failure includes various TCM syndrome texts for chronic heart failure, including Qi and Yin deficiency syndrome, Qi deficiency and blood stasis syndrome, heart and kidney Yang deficiency syndrome, and Yang deficiency and water retention syndrome.
[0006] Optionally, the first mapping relationship is used as a label to annotate the TCM syndrome type text set of chronic heart failure, resulting in the first-labeled TCM syndrome type text set of chronic heart failure, including: The first mapping relationship includes the SDNN parameter range and rMSSD parameter range corresponding to Qi and Yin deficiency syndrome, Qi deficiency and blood stasis syndrome, heart and kidney Yang deficiency syndrome, and Yang deficiency and water retention syndrome, respectively. The first mapping relationship is used as a label to label Qi and Yin deficiency syndrome, Qi deficiency and blood stasis syndrome, heart and kidney Yang deficiency syndrome and Yang deficiency and water retention syndrome, and the first label Qi and Yin deficiency syndrome, the first label Qi deficiency and blood stasis syndrome, the first label heart and kidney Yang deficiency syndrome and the first label Yang deficiency and water retention syndrome are obtained. The first label consists of the TCM syndrome text set of chronic heart failure, which includes the first label Qi and Yin deficiency syndrome, the first label Qi deficiency and blood stasis syndrome, the first label heart and kidney Yang deficiency syndrome, and the first label Yang deficiency and water retention syndrome.
[0007] Optionally, the second mapping relationship is used as a label to annotate the TCM syndrome type text set of chronic heart failure, resulting in a second-labeled TCM syndrome type text set of chronic heart failure, including: The second mapping relationship includes the LF parameter range, HF parameter range, and LF / HF parameter range corresponding to Qi and Yin deficiency syndrome, Qi deficiency and blood stasis syndrome, heart and kidney Yang deficiency syndrome, and Yang deficiency and water retention syndrome, respectively. The second mapping relationship is used as a label to annotate the Qi and Yin deficiency syndrome, Qi deficiency and blood stasis syndrome, heart and kidney Yang deficiency syndrome and Yang deficiency and water retention syndrome, and the second label Qi and Yin deficiency syndrome, the second label Qi deficiency and blood stasis syndrome, the second label heart and kidney Yang deficiency syndrome and the second label Yang deficiency and water retention syndrome are obtained. The second label consists of the TCM syndrome text set of chronic heart failure, which includes the second label Qi and Yin deficiency syndrome, the second label Qi deficiency and blood stasis syndrome, the second label heart and kidney Yang deficiency syndrome, and the second label Yang deficiency and water retention syndrome.
[0008] Optionally, the first initial TCM syndrome identification model includes a first CNN module, an LSTM module, a first fully connected module, and a first Softmax module; The first initial TCM syndrome identification model was iteratively trained using a time-domain dataset and a first-label TCM syndrome type text set for chronic heart failure. After training, the first final TCM syndrome identification model was obtained, including: S11: Input the i-th time-domain data group in the time-domain dataset into the first CNN module, extract features from the SDNN and rMSSD parameters to obtain SDNN features and rMSSD features, and concatenate the SDNN features and rMSSD features to form the initial time-domain feature vector. ; S12: Initial time-domain feature vector The input LSTM module captures the dependencies in the features to obtain the temporal hidden state feature vector. ; transform the temporal hidden state feature vector The input is processed by the first fully connected module to perform feature dimensionality reduction and nonlinear transformation, resulting in the final time-domain feature vector. ; S13: The final time-domain feature vector Input the first Softmax module to perform probability prediction and obtain the first predicted probability distribution. Take the text of chronic heart failure TCM syndrome with the highest probability in the first predicted probability distribution as the first predicted chronic heart failure TCM syndrome text. S14: In the first-label chronic heart failure TCM syndrome type text set, obtain the first-label chronic heart failure TCM syndrome type text corresponding to the i-th time domain data group; calculate the first loss function value based on the probability of the first-label chronic heart failure TCM syndrome type text and the probability of the first predicted chronic heart failure TCM syndrome type text. S15: Backpropagate the first initial TCM syndrome identification model based on the first loss function value and adjust the parameters of the first initial TCM syndrome identification model. S16: Repeat steps S11-S15 until the first loss function value satisfies the first convergence condition, and obtain the first final TCM syndrome identification model.
[0009] Optionally, the second initial TCM syndrome identification model includes a second CNN module, a Transformer module, a second fully connected module, and a second Softmax module; The second initial TCM syndrome identification model is iteratively trained using a frequency domain dataset and a second-label TCM syndrome text set for chronic heart failure. After training, the second final TCM syndrome identification model is obtained, which includes: S21: Input the j-th frequency domain data group from the frequency domain data set into the second CNN module, extract features from the LF parameters, HF parameters, and LF / HF parameters to obtain LF features, HF features, and LF / HF features, and concatenate the LF features, HF features, and LF / HF features into an initial frequency domain feature vector. ; S22: Initial frequency domain feature vector The input is encoded and decoded by the Transformer module to obtain the frequency domain hidden state feature vector. The frequency domain hidden state feature vector The input is processed by the second fully connected module to perform feature dimensionality reduction and nonlinear transformation, resulting in the final frequency domain feature vector. ; S23: The final frequency domain feature vector Input the second Softmax module to perform probability prediction and obtain the second predicted probability distribution. The text of the chronic heart failure TCM syndrome with the highest probability in the second predicted probability distribution is taken as the second predicted chronic heart failure TCM syndrome text. S24: In the set of TCM syndrome type texts for chronic heart failure with second label, obtain the TCM syndrome type text for chronic heart failure with second label corresponding to the j-th frequency domain data group; calculate the second loss function value based on the probability of the TCM syndrome type text for chronic heart failure with second label and the probability of the second predicted TCM syndrome type text for chronic heart failure. S25: Backpropagate the second initial TCM syndrome identification model based on the second loss function value and adjust the parameters of the second initial TCM syndrome identification model. S26: Repeat steps S21-S25 until the value of the second loss function satisfies the second convergence condition, and obtain the second final TCM syndrome identification model.
[0010] Optionally, the probability sets of the first and second TCM syndrome types are weighted and summed, and the maximum probability is selected to obtain the final TCM syndrome type probability. Based on the final TCM syndrome type probability, the final TCM syndrome type text for chronic heart failure is obtained, including: The first set of probabilities for TCM syndrome types includes the probabilities of the first Qi and Yin deficiency syndrome, the first Qi deficiency and blood stasis syndrome, the first heart and kidney Yang deficiency syndrome, and the first Yang deficiency and water retention syndrome. The second set of probabilities for TCM syndrome types includes the probabilities of the second Qi and Yin deficiency syndrome, the second Qi deficiency and blood stasis syndrome, the second heart and kidney Yang deficiency syndrome, and the second Yang deficiency and water retention syndrome. The probability of the first Qi and Yin deficiency syndrome is obtained by weighted summing the probabilities of the second Qi and Yin deficiency syndrome; the probability of the first Qi deficiency and blood stasis syndrome is obtained by weighted summing the probabilities of the second Qi deficiency and blood stasis syndrome; the probability of the first heart and kidney Yang deficiency syndrome is obtained by weighted summing the probabilities of the second heart and kidney Yang deficiency syndrome; the probability of the first Yang deficiency and water retention syndrome is obtained by weighted summing the probabilities of the second Yang deficiency and water retention syndrome. The highest probability among the probabilities of Qi and Yin deficiency syndrome, Qi deficiency and blood stasis syndrome, heart and kidney Yang deficiency syndrome, and Yang deficiency and water retention syndrome is taken as the final TCM syndrome probability. The TCM syndrome text of chronic heart failure corresponding to the final TCM syndrome probability is taken as the final TCM syndrome text of chronic heart failure.
[0011] The present invention also provides a TCM syndrome differentiation device for chronic heart failure, used to implement the aforementioned TCM syndrome differentiation method for chronic heart failure, the device comprising: The data acquisition module is used to acquire a heart rate variability dataset and a TCM syndrome type text set for chronic heart failure. The heart rate variability dataset includes a time domain dataset and a frequency domain dataset. The time-domain annotation module is used to obtain the first mapping relationship between the time-domain data set and the TCM syndrome type text set of chronic heart failure. The first mapping relationship is used as a label to annotate the TCM syndrome type text set of chronic heart failure, and the first label chronic heart failure TCM syndrome type text set is obtained. The frequency domain labeling module is used to obtain the second mapping relationship between the frequency domain data set and the TCM syndrome type text set of chronic heart failure. The second mapping relationship is used as a label to label the TCM syndrome type text set of chronic heart failure, and the second label chronic heart failure TCM syndrome type text set is obtained. The first initial TCM syndrome identification model training module is used to iteratively train the first initial TCM syndrome identification model using a time-domain data set and a first-label chronic heart failure TCM syndrome text set. After training, the first final TCM syndrome identification model is obtained. The second initial TCM syndrome identification model training module is used to iteratively train the second initial TCM syndrome identification model using a frequency domain data set and a second-label chronic heart failure TCM syndrome text set. After training, the second final TCM syndrome identification model is obtained. The TCM syndrome probability acquisition module is used to acquire the heart rate variability data set to be tested, divide the heart rate variability data set to be tested into a time domain data set to be tested and a frequency domain data set to be tested, input the time domain data set to be tested into a first final TCM syndrome identification model for TCM syndrome prediction, and obtain a first TCM syndrome probability set; input the frequency domain data set to be tested into a second final TCM syndrome identification model for TCM syndrome prediction, and obtain a second TCM syndrome probability set. The final TCM syndrome type text acquisition module for chronic heart failure is used to perform a weighted summation of the first TCM syndrome type probability set and the second TCM syndrome type probability set and select the maximum probability to obtain the final TCM syndrome type probability, and obtain the final TCM syndrome type text for chronic heart failure based on the final TCM syndrome type probability.
[0012] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method for identifying TCM syndrome types of chronic heart failure.
[0013] The present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method for identifying TCM syndrome types of chronic heart failure.
[0014] The present invention has the following beneficial effects: 1. Heart rate variability data was divided into time-domain and frequency-domain data. Time-domain data reflects the overall activity of the autonomic nervous system, while frequency-domain data distinguishes the balance between the sympathetic and parasympathetic nervous systems. The TCM syndrome texts for chronic heart failure were labeled and the model trained using both time-domain and frequency-domain data, overcoming the limitations of single-modal data. The analysis of multimodal data improved the recognition accuracy of TCM syndrome texts. By weighted summation of the first and second TCM syndrome probability sets and selecting the maximum probability, the final TCM syndrome text for chronic heart failure was obtained. This method can automatically correct misjudgments by a single model. When one model is affected by noise, another model can correct the bias, further improving the prediction accuracy of TCM syndrome texts for chronic heart failure. 2. A first initial TCM syndrome identification model is constructed using the LSTM module. The LSTM module can capture the temporal dependency of heart rate variability parameters in the time domain data. The first initial TCM syndrome identification model is iteratively trained using the time domain data set and the first label TCM syndrome text set of chronic heart failure. This allows the first initial TCM syndrome identification model to focus on identifying TCM syndrome text of chronic heart failure through time domain data, thereby improving the accuracy of TCM syndrome text prediction through time domain data. 3. A second initial TCM syndrome identification model is constructed using the Transformer module. The Transformer module can capture the global dependencies of the frequency domain data of heart rate variability parameters. The second initial TCM syndrome identification model is iteratively trained using the frequency domain data set and the second-label chronic heart failure TCM syndrome text set, so that the second initial TCM syndrome identification model focuses on identifying chronic heart failure TCM syndrome text through frequency domain data, thereby improving the accuracy of TCM syndrome text prediction through frequency domain data. Attached Figure Description
[0015] Figure 1 This is a flowchart of a method according to an embodiment of the present invention; Figure 2 This is a structural diagram of the device according to an embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation
[0016] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.
[0017] Reference Figure 1 This invention provides a method for identifying TCM syndrome types in chronic heart failure, comprising the following steps: Acquire a set of heart rate variability data and a set of TCM syndrome types for chronic heart failure. The heart rate variability data set includes a time domain data set and a frequency domain data set. In some embodiments, the time-domain dataset includes multiple time-domain data groups, and one time-domain data group includes an SDNN parameter and an rMSSD parameter; The frequency domain dataset consists of multiple frequency domain data sets. Each frequency domain data set includes an LF parameter, an HF parameter, and an LF / HF parameter. The collection of TCM syndrome texts for chronic heart failure includes various TCM syndrome texts for chronic heart failure, including Qi and Yin deficiency syndrome, Qi deficiency and blood stasis syndrome, heart and kidney Yang deficiency syndrome, and Yang deficiency and water retention syndrome.
[0018] In some embodiments, time-domain data uses SDNN parameters and rMSSD parameters, while frequency-domain data uses LF parameters, HF parameters, and LF / HF parameters. SDNN parameters represent the standard deviation of normal sinus intervals, used to assess the regulatory capacity of the autonomic nervous system (sympathetic and parasympathetic) on heart rhythm, measured in milliseconds (ms). rMSSD parameters represent the root mean square of the difference between adjacent normal intervals, used to quantify the instantaneous variation in heart rate intervals, reflecting the instantaneous regulatory capacity of the parasympathetic (vagus) nervous system on the heart, measured in milliseconds (ms). LF parameters represent low-frequency power, measured in power spectral density (ms). 2 This reflects the combined effects of the sympathetic and vagus nerves; the HF parameter represents high-frequency power, measured in power spectral density (ms). 2 The LF / HF parameter represents the change in vagal tone; it represents the ratio of low-frequency power to high-frequency power and is used to assess the balance between the sympathetic and vagal nerves.
[0019] Qi and Yin deficiency syndrome is mainly manifested as qi deficiency and fatigue, and yin deficiency and dryness, such as shortness of breath, reluctance to speak, fatigue, dry mouth and throat, and a thready and rapid pulse. It is often treated with qi-tonifying and yin-nourishing formulas such as Huangqi Shengmai San. Qi deficiency and blood stasis syndrome is characterized by the coexistence of qi deficiency and weakness (such as shortness of breath and spontaneous sweating) and blood stasis (such as chest tightness and stabbing pain, and a purplish tongue). It is often seen after prolonged illness or chronic diseases. Treatment requires tonifying qi and promoting blood circulation, and formulas such as Taoren Honghua Jian are used. Heart and kidney yang deficiency syndrome is characterized by symptoms of heart yang deficiency and cold, such as cold limbs, palpitations, and aversion to cold. Kidney yang deficiency aggravates cold pain in the lower back and knees. The treatment is mainly to warm and tonify the heart and kidneys. Shenfu Tang combined with Wuling San can warm yang and promote diuresis. Yang deficiency and water retention syndrome is caused by yang deficiency and water retention leading to upward overflow of water evil. Common symptoms include cough and pink frothy sputum. It requires draining the lungs and promoting water metabolism. Zhenwu Tang plus Tingli Dazao Xiefei Tang can warm kidney yang and drain water.
[0020] Obtain the first mapping relationship between the time-domain data set and the TCM syndrome type text set of chronic heart failure, and use the first mapping relationship as a label to annotate the TCM syndrome type text set of chronic heart failure to obtain the first label chronic heart failure TCM syndrome type text set. In some embodiments, the first mapping relationship is used as a label to annotate the TCM syndrome type text set of chronic heart failure, resulting in a first-labeled TCM syndrome type text set of chronic heart failure, including: The first mapping relationship includes the SDNN parameter range and rMSSD parameter range corresponding to Qi and Yin deficiency syndrome, Qi deficiency and blood stasis syndrome, heart and kidney Yang deficiency syndrome, and Yang deficiency and water retention syndrome, respectively. The first mapping relationship is used as a label to label Qi and Yin deficiency syndrome, Qi deficiency and blood stasis syndrome, heart and kidney Yang deficiency syndrome and Yang deficiency and water retention syndrome, and the first label Qi and Yin deficiency syndrome, the first label Qi deficiency and blood stasis syndrome, the first label heart and kidney Yang deficiency syndrome and the first label Yang deficiency and water retention syndrome are obtained. The first label consists of the TCM syndrome text set of chronic heart failure, which includes the first label Qi and Yin deficiency syndrome, the first label Qi deficiency and blood stasis syndrome, the first label heart and kidney Yang deficiency syndrome, and the first label Yang deficiency and water retention syndrome.
[0021] In some embodiments, an objective correspondence between TCM syndrome types and autonomic nervous system function is established based on temporal data of heart rate variability (SDNN parameters, rMSSD parameters). Specific SDNN and rMSSD parameter ranges are set according to the patient's case, and these ranges require symptomatological cross-validation by a physician. For example, the SDNN parameter range corresponding to the first label, Qi and Yin deficiency syndrome, is 80-120ms, and the rMSSD parameter range is 15-30ms, with clinical features including shortness of breath, wheezing, and Yin deficiency symptoms. The SDNN parameter range corresponding to the first label, Qi deficiency and blood stasis syndrome, is 60-80ms, and the rMSSD parameter range is 10-15ms, with clinical features including fatigue, palpitations, and blood stasis in the tongue and pulse. The SDNN parameter range corresponding to the first label, Heart and Kidney Yang deficiency syndrome, is 40-60ms, and the rMSSD parameter range is 5-10ms, with clinical features including aversion to cold, cold limbs, and Yang deficiency manifestations. The first label, Yang deficiency with water retention, corresponds to SDNN parameters in the range of <40ms and rMSSD parameters in the range of <5ms. Its clinical features are edema, shortness of breath, and signs of water retention.
[0022] Obtain the second mapping relationship between the frequency domain data set and the TCM syndrome type text set of chronic heart failure. Use the second mapping relationship as a label to annotate the TCM syndrome type text set of chronic heart failure, and obtain the second label TCM syndrome type text set of chronic heart failure. In some embodiments, the second mapping relationship is used as a label to annotate the TCM syndrome type text set of chronic heart failure, resulting in a second-labeled TCM syndrome type text set of chronic heart failure, including: The second mapping relationship includes the LF parameter range, HF parameter range, and LF / HF parameter range corresponding to Qi and Yin deficiency syndrome, Qi deficiency and blood stasis syndrome, heart and kidney Yang deficiency syndrome, and Yang deficiency and water retention syndrome, respectively. The second mapping relationship is used as a label to annotate the Qi and Yin deficiency syndrome, Qi deficiency and blood stasis syndrome, heart and kidney Yang deficiency syndrome and Yang deficiency and water retention syndrome, and the second label Qi and Yin deficiency syndrome, the second label Qi deficiency and blood stasis syndrome, the second label heart and kidney Yang deficiency syndrome and the second label Yang deficiency and water retention syndrome are obtained. The second label consists of the TCM syndrome text set of chronic heart failure, which includes the second label Qi and Yin deficiency syndrome, the second label Qi deficiency and blood stasis syndrome, the second label heart and kidney Yang deficiency syndrome, and the second label Yang deficiency and water retention syndrome.
[0023] In some embodiments, an objective correspondence between TCM syndrome types and autonomic nervous system function is established based on frequency domain data of heart rate variability (low-frequency power LF, high-frequency power HF, and LF / HF ratio). Specific ranges for LF, HF, and LF / HF parameters are set according to the patient's case. These ranges require symptomatological cross-validation by a physician. For example, the LF parameter range corresponding to the second label, Qi and Yin deficiency syndrome, is 400-800 ms. 2 The HF parameter range is 150-300 ms. 2 The LF / HF ratio ranged from 2.5 to 4.0. The LF parameter range for the second label, Qi deficiency and blood stasis syndrome, was 200-500 ms. 2 The HF parameter range is 100-250 ms. 2 The LF / HF ratio ranged from 1.2 to 2.3. The LF parameter range for the second label, Heart and Kidney Yang Deficiency Syndrome, was 50-300 ms. 2 The HF parameter range is 50-150 ms. 2 The LF / HF ratio ranges from 0.8 to 1.5. For the second label, Yang deficiency with water retention syndrome, the LF parameter ranges from 100 to 400 ms. 2 The HF parameter range is 30-100 ms. 2 The LF / HF ratio ranges from 3.0 to 6.0.
[0024] The first initial TCM syndrome identification model is iteratively trained using a time-domain dataset and a first-label TCM syndrome text set for chronic heart failure. After training, the first final TCM syndrome identification model is obtained. In some embodiments, the first initial TCM syndrome identification model includes a first CNN module, an LSTM module, a first fully connected module, and a first Softmax module; In some embodiments, the first CNN module employs a dual-channel design, with each channel having a 1D convolutional kernel to capture features of the input values. The LSTM module uses a Long Short-Term Memory (LSTM) network, which effectively learns long-term dependencies in sequence data by introducing cell states (long-term memory channels) and gating mechanisms (forget gate, input gate, and output gate). The cell state is the LSTM's "memory channel," where information can flow relatively invariantly over a long period. The gating mechanisms are the "switches" that control the flow of information (inflow, storage, and outflow from the cell state), implemented by Sigmoid neural network layers (outputting values from 0 to 1) and pointwise multiplication operations. The forget gate determines which information is discarded from the cell state, the input gate determines which new information is stored in the cell state, and the output gate determines what to output based on the current cell state.
[0025] The first initial TCM syndrome identification model was iteratively trained using a time-domain dataset and a first-label TCM syndrome type text set for chronic heart failure. After training, the first final TCM syndrome identification model was obtained, including: S11: Input the i-th time-domain data group in the time-domain dataset into the first CNN module, extract features from the SDNN and rMSSD parameters to obtain SDNN features and rMSSD features, and concatenate the SDNN features and rMSSD features to form the initial time-domain feature vector. ; In some embodiments, SDNN features are obtained by performing convolution operations on the SDNN parameters through the first channel of the first CNN module. The rMSSD features are obtained by performing a convolution operation on the second channel of the first CNN module. SDNN features are concatenated using a concatenation function. and rMSSD features Concatenate to form the initial time-domain feature vector .
[0026] S12: Initial time-domain feature vector The input LSTM module captures the dependencies in the features to obtain the temporal hidden state feature vector. ; transform the temporal hidden state feature vector The input is processed by the first fully connected module to perform feature dimensionality reduction and nonlinear transformation, resulting in the final time-domain feature vector. ; In some embodiments, the initial temporal feature vector is obtained. eigenvectors at time step t The feature vector of the forget gate at time step t The expression is: in, To adjust the parameters, As for the weight of the Forgotten Gate, For the bias parameters of the forget gate, Let be the hidden state feature vector at time step t-1; The feature vector of the input gate at time step t and cell state The expression is: in, The weights of the input gates, These are the bias parameters for the input gate. Weights for cell states, This is a bias parameter for the cell state.
[0027] The expression for updating cell state is: C t =x t ⊙ C t-1 +i t ⊙ Here, ⊙ represents the XOR operation.
[0028] eigenvectors of the output gate The expression is: in, The weights of the output gates, These are the bias parameters for the output gate.
[0029] Hidden state feature vector at time step t The expression is: h t =o t ⊙ tanh(C t ) ; Temporal hidden state feature vector The expression is: ,in This is the hidden state feature vector at the last time step.
[0030] Select The input is processed by the first fully connected module for feature dimensionality reduction and nonlinear transformation, resulting in the final time-domain feature vector. The expression is: in, Represents the ReLU activation function. As the first weight, This is the first bias parameter.
[0031] S13: The final time-domain feature vector Input the first Softmax module to perform probability prediction and obtain the first predicted probability distribution. Take the text of chronic heart failure TCM syndrome with the highest probability in the first predicted probability distribution as the first predicted chronic heart failure TCM syndrome text. S14: In the first-label chronic heart failure TCM syndrome type text set, obtain the first-label chronic heart failure TCM syndrome type text corresponding to the i-th time domain data group; calculate the first loss function value based on the probability of the first-label chronic heart failure TCM syndrome type text and the probability of the first predicted chronic heart failure TCM syndrome type text. In some embodiments, if the SDNN parameters and rMSSD parameters of the i-th time-domain data group are both within the first-label chronic heart failure TCM syndrome text set, and the SDNN parameter range and rMSSD parameter range of a certain TCM syndrome text are within the range, then the TCM syndrome text is taken as the first-label chronic heart failure TCM syndrome text corresponding to the i-th time-domain data group, and the probability of the first-label chronic heart failure TCM syndrome text is obtained from the historical training database; otherwise, the i-th time-domain data group is determined to be invalid, and S11 is returned to obtain a new time-domain data group.
[0032] The formula for calculating the first loss function value L1 is: in, The probability of a text labeled as a TCM syndrome type with the first tag being chronic heart failure is given, where j ranges from 1 to 4, corresponding to the four TCM syndrome types respectively. Let i be the probability of the first text predicting the TCM syndrome type of chronic heart failure, where i ranges from 1 to 4, corresponding to the four TCM syndrome types respectively. This is the first adjustment parameter. Weights among TCM syndrome types of chronic heart failure.
[0033] S15: Backpropagate the first initial TCM syndrome identification model based on the first loss function value and adjust the parameters of the first initial TCM syndrome identification model. S16: Repeat steps S11-S15 until the first loss function value satisfies the first convergence condition, and obtain the first final TCM syndrome identification model.
[0034] The second initial TCM syndrome identification model is iteratively trained using a frequency domain dataset and a second-label TCM syndrome text set for chronic heart failure. After training, the second final TCM syndrome identification model is obtained. In some embodiments, the second initial TCM syndrome identification model includes a second CNN module, a Transformer module, a second fully connected module, and a second Softmax module; In some embodiments, the second CNN module employs a three-channel design, with each channel having a 1D convolutional kernel to capture features of the input values. The Transformer module uses an encoder-decoder structure, where the encoder consists of N identical stacked layers, each containing a multi-head attention mechanism module, a feedforward neural network, residual connections, and layer normalization. The decoder also has multiple stacked layers, each containing a mask self-attention module and an encoder-decoder attention module. The feature vectors input to the multi-head attention mechanism module are transformed in parallel to generate multiple independent sets of query vector (Q), key vector (K), and value vector (V) matrices, each set being called a "head". Each head computes attention in an independent subspace, avoiding feature omissions caused by the single-head mechanism over-focusing on its own position. The feedforward neural network is used to perform position-wise nonlinear transformations on the feature vectors. Residual connections and layer normalization are used to stabilize training and accelerate convergence. The mask self-attention module is used to prevent the model from "peeping" at future information during task generation, ensuring that predictions only depend on the generated content. The encoder-decoder attention module is used to associate the encoder output with the decoder input.
[0035] The second initial TCM syndrome identification model is iteratively trained using a frequency domain dataset and a second-label TCM syndrome text set for chronic heart failure. After training, the second final TCM syndrome identification model is obtained, which includes: S21: Input the j-th frequency domain data group from the frequency domain data set into the second CNN module, extract features from the LF parameters, HF parameters, and LF / HF parameters to obtain LF features, HF features, and LF / HF features, and concatenate the LF features, HF features, and LF / HF features into an initial frequency domain feature vector. ; In some embodiments, LF features are obtained by performing convolution operations on the LF parameters through the first channel of the second CNN module. The HF features are obtained by performing convolution operations on the HF parameters through the second channel of the second CNN module. The LF / HF features are obtained by performing convolution operations on the LF / HF parameters through the third channel of the second CNN module. LF features are concatenated using a concatenation function. HF characteristics and LF / HF characteristics Concatenate to form the initial frequency domain feature vector .
[0036] S22: Initial frequency domain feature vector The input is encoded and decoded by the Transformer module to obtain the frequency domain hidden state feature vector. The frequency domain hidden state feature vector The input is processed by the second fully connected module to perform feature dimensionality reduction and nonlinear transformation, resulting in the final frequency domain feature vector. ; In some embodiments, the initial frequency domain feature vector The input encoder is used to obtain the multi-head output feature vector through multi-head self-attention computation. The expression is: in, This represents the attention weight of the k-th self-attention head; Multi-head output feature vector The input is decoded by the decoder to obtain the frequency domain hidden state feature vector. The expression is: in, Presentation layer normalization operation.
[0037] Hidden state feature vector in the frequency domain The input is processed by the second fully connected module for feature dimensionality reduction and nonlinear transformation, resulting in the final frequency domain feature vector. The expression is: in, Represents the ReLU activation function. As the second weight, This is the second bias parameter.
[0038] S23: The final frequency domain feature vector Input the second Softmax module to perform probability prediction and obtain the second predicted probability distribution. The text of the chronic heart failure TCM syndrome with the highest probability in the second predicted probability distribution is taken as the second predicted chronic heart failure TCM syndrome text. S24: In the set of TCM syndrome type texts for chronic heart failure with second label, obtain the TCM syndrome type text for chronic heart failure with second label corresponding to the j-th frequency domain data group; calculate the second loss function value based on the probability of the TCM syndrome type text for chronic heart failure with second label and the probability of the second predicted TCM syndrome type text for chronic heart failure. In some embodiments, if the LF parameter, HF parameter, and LF / HF parameter of the j-th frequency domain data group are all within the second-label chronic heart failure TCM syndrome text set, and the LF parameter range, HF parameter range, and LF / HF parameter range of a certain TCM syndrome text are within the range of a certain TCM syndrome text, then the TCM syndrome text is taken as the second-label chronic heart failure TCM syndrome text corresponding to the j-th frequency domain data group, and the probability of the second-label chronic heart failure TCM syndrome text is obtained from the historical training database; otherwise, the j-th frequency domain data group is determined to be invalid, and the process returns to S21 to obtain a new frequency domain data group.
[0039] The formula for calculating the second loss function value L2 is: in, The probability of a text labeled "chronic heart failure" as a TCM syndrome type is given by the second label. u ranges from 1 to 4, corresponding to four different TCM syndrome types. The probability of predicting the TCM syndrome type of chronic heart failure in the second text is given by v, which ranges from 1 to 4, corresponding to the four TCM syndrome types respectively. This is the second adjustment parameter. Weights among TCM syndrome types of chronic heart failure.
[0040] S25: Backpropagate the second initial TCM syndrome identification model based on the second loss function value and adjust the parameters of the second initial TCM syndrome identification model. S26: Repeat steps S21-S25 until the value of the second loss function satisfies the second convergence condition, and obtain the second final TCM syndrome identification model.
[0041] A set of heart rate variability data to be tested is obtained, and the set of heart rate variability data to be tested is divided into a time domain data set and a frequency domain data set to be tested. The time domain data set is input into the first final TCM syndrome identification model for TCM syndrome prediction to obtain the first TCM syndrome probability set; the frequency domain data set is input into the second final TCM syndrome identification model for TCM syndrome prediction to obtain the second TCM syndrome probability set. The first set of TCM syndrome probabilities and the second set of TCM syndrome probabilities are weighted and summed, and the maximum probability is selected to obtain the final TCM syndrome probability. The final TCM syndrome text of chronic heart failure is obtained based on the final TCM syndrome probability.
[0042] In some embodiments, the probability sets of the first and second TCM syndrome types are weighted and summed, and the maximum probability is selected to obtain the final TCM syndrome type probability. Based on the final TCM syndrome type probability, the final TCM syndrome type text for chronic heart failure is obtained, including: The first set of probabilities for TCM syndrome types includes the probabilities of the first Qi and Yin deficiency syndrome, the first Qi deficiency and blood stasis syndrome, the first heart and kidney Yang deficiency syndrome, and the first Yang deficiency and water retention syndrome. The second set of probabilities for TCM syndrome types includes the probabilities of the second Qi and Yin deficiency syndrome, the second Qi deficiency and blood stasis syndrome, the second heart and kidney Yang deficiency syndrome, and the second Yang deficiency and water retention syndrome. The probability of the first Qi and Yin deficiency syndrome is obtained by weighted summing the probabilities of the second Qi and Yin deficiency syndrome; the probability of the first Qi deficiency and blood stasis syndrome is obtained by weighted summing the probabilities of the second Qi deficiency and blood stasis syndrome; the probability of the first heart and kidney Yang deficiency syndrome is obtained by weighted summing the probabilities of the second heart and kidney Yang deficiency syndrome; the probability of the first Yang deficiency and water retention syndrome is obtained by weighted summing the probabilities of the second Yang deficiency and water retention syndrome. The highest probability among the probabilities of Qi and Yin deficiency syndrome, Qi deficiency and blood stasis syndrome, heart and kidney Yang deficiency syndrome, and Yang deficiency and water retention syndrome is taken as the final TCM syndrome probability. The TCM syndrome text of chronic heart failure corresponding to the final TCM syndrome probability is taken as the final TCM syndrome text of chronic heart failure.
[0043] In some embodiments, the probabilities of the first Qi and Yin deficiency syndrome, the first Qi deficiency and blood stasis syndrome, the first heart and kidney Yang deficiency syndrome, and the first Yang deficiency and water retention syndrome are respectively p 11 p 12 p 13 and p 14 The probabilities of the second Qi and Yin deficiency syndrome, the second Qi deficiency and blood stasis syndrome, the second heart and kidney Yang deficiency syndrome, and the second Yang deficiency and water retention syndrome are respectively p 21 p 22 p 23 and p 24 The probabilities of merging Qi and Yin deficiency syndrome, Qi deficiency and blood stasis syndrome, heart and kidney Yang deficiency syndrome, and Yang deficiency with water retention syndrome are respectively W1×p 11 +W2×p 21 W3×p 12 +W4×p 22 W5×p 13 +W6×p 23 and W7×p 14 +W8×p 24 Where W1, W2, W3, W4, W5, W6, W7, and W8 are all weight parameters, if W1 × p 11 +W2×p 21 The most extreme case is the syndrome of deficiency of both Qi and Yin, which is considered the final TCM syndrome type for chronic heart failure.
[0044] The present invention also provides a TCM syndrome differentiation device 20 for chronic heart failure, used to implement the aforementioned TCM syndrome differentiation method for chronic heart failure, the device comprising: The data acquisition module 21 is used to acquire a heart rate variability dataset and a TCM syndrome type text set for chronic heart failure. The heart rate variability dataset includes a time domain dataset and a frequency domain dataset. The time-domain annotation module 22 is used to obtain the first mapping relationship between the time-domain data set and the TCM syndrome text set of chronic heart failure, and to use the first mapping relationship as a label to annotate the TCM syndrome text set of chronic heart failure, thereby obtaining the first label TCM syndrome text set of chronic heart failure. The frequency domain labeling module 23 is used to obtain the second mapping relationship between the frequency domain data set and the TCM syndrome text set of chronic heart failure, and to use the second mapping relationship as a label to label the TCM syndrome text set of chronic heart failure, thereby obtaining the second label TCM syndrome text set of chronic heart failure. The first initial TCM syndrome identification model training module 24 is used to iteratively train the first initial TCM syndrome identification model using a time-domain data set and a first-label chronic heart failure TCM syndrome text set, and obtain the first final TCM syndrome identification model after training. The second initial TCM syndrome identification model training module 25 is used to iteratively train the second initial TCM syndrome identification model through a frequency domain data set and a second-label chronic heart failure TCM syndrome text set, and obtain the second final TCM syndrome identification model after training. The TCM syndrome probability acquisition module 26 is used to acquire the heart rate variability data set to be tested, divide the heart rate variability data set to be tested into a time domain data set to be tested and a frequency domain data set to be tested, input the time domain data set to be tested into the first final TCM syndrome identification model for TCM syndrome prediction, and obtain the first TCM syndrome probability set; input the frequency domain data set to be tested into the second final TCM syndrome identification model for TCM syndrome prediction, and obtain the second TCM syndrome probability set. The final TCM syndrome type text acquisition module 27 for chronic heart failure is used to perform weighted summation and maximum probability selection of the first TCM syndrome type probability set and the second TCM syndrome type probability set to obtain the final TCM syndrome type probability, and obtain the final TCM syndrome type text for chronic heart failure based on the final TCM syndrome type probability.
[0045] This application provides an electronic device, including a processor and a memory; the memory stores a computer program, wherein the computer program, when executed by the processor, implements any of the above-described schemes of the method for identifying TCM syndrome types of chronic heart failure.
[0046] Specifically, the processor may include, for example, a general-purpose microprocessor, an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor may also include onboard memory for caching purposes. The processor may be a single processing unit or multiple processing units for performing different actions of the method flow according to embodiments of this application.
[0047] Memory can be any medium capable of containing, storing, transmitting, propagating, or transmitting instructions. For example, memory can include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, instruments, or propagation media. Specific examples of memory include: magnetic storage devices such as magnetic tape or hard disk drives (HDDs); optical storage devices such as optical discs (CD-ROMs); and also random access memory (RAM) or flash memory; and / or wired / wireless communication links.
[0048] This application also provides a computer-readable medium storing a computer program thereon, which, when executed by a processor, implements the method for identifying TCM syndrome types in chronic heart failure according to any of the above-described schemes. This computer-readable medium may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into that device / apparatus / system. The aforementioned computer-readable medium carries one or more programs, which, when executed, implement the method as described in the embodiments of this application.
[0049] According to embodiments of this application, a computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wired, optical fiber, radio frequency signals, etc., or any suitable combination thereof.
[0050] Those skilled in the art will understand that the features described in the various embodiments and / or claims of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments and / or claims of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application. Therefore, the scope of this application should not be limited to the above embodiments, but should be defined not only by the appended claims, but also by their equivalents. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.
Claims
1. A method for identifying TCM syndrome types in chronic heart failure, characterized in that, Including the following steps: Acquire a set of heart rate variability data and a set of TCM syndrome types for chronic heart failure. The heart rate variability data set includes a time domain data set and a frequency domain data set. Obtain the first mapping relationship between the time-domain data set and the TCM syndrome type text set of chronic heart failure, and use the first mapping relationship as a label to annotate the TCM syndrome type text set of chronic heart failure to obtain the first label chronic heart failure TCM syndrome type text set. Obtain the second mapping relationship between the frequency domain data set and the TCM syndrome type text set of chronic heart failure. Use the second mapping relationship as a label to annotate the TCM syndrome type text set of chronic heart failure, and obtain the second label TCM syndrome type text set of chronic heart failure. The first initial TCM syndrome identification model is iteratively trained using a time-domain dataset and a first-label TCM syndrome text set for chronic heart failure. After training, the first final TCM syndrome identification model is obtained. The second initial TCM syndrome identification model is iteratively trained using a frequency domain dataset and a second-label TCM syndrome text set for chronic heart failure. After training, the second final TCM syndrome identification model is obtained. A set of heart rate variability data to be tested is obtained, and the set of heart rate variability data to be tested is divided into a time domain data set and a frequency domain data set to be tested. The time domain data set is input into the first final TCM syndrome identification model for TCM syndrome prediction to obtain the first TCM syndrome probability set; the frequency domain data set is input into the second final TCM syndrome identification model for TCM syndrome prediction to obtain the second TCM syndrome probability set. The first set of TCM syndrome probabilities and the second set of TCM syndrome probabilities are weighted and summed, and the maximum probability is selected to obtain the final TCM syndrome probability. The final TCM syndrome text of chronic heart failure is obtained based on the final TCM syndrome probability.
2. The method for identifying TCM syndrome types in chronic heart failure according to claim 1, characterized in that: The time-domain dataset consists of multiple time-domain data sets, and each time-domain data set includes one SDNN parameter and one rMSSD parameter. The frequency domain dataset consists of multiple frequency domain data sets. Each frequency domain data set includes an LF parameter, an HF parameter, and an LF / HF parameter. The collection of TCM syndrome texts for chronic heart failure includes various TCM syndrome texts for chronic heart failure, including Qi and Yin deficiency syndrome, Qi deficiency and blood stasis syndrome, heart and kidney Yang deficiency syndrome, and Yang deficiency and water retention syndrome.
3. The method for identifying TCM syndrome types in chronic heart failure according to claim 2, characterized in that, The first mapping relationship is used as a label to annotate the TCM syndrome type text set of chronic heart failure, resulting in the first-labeled TCM syndrome type text set of chronic heart failure, including: The first mapping relationship includes the SDNN parameter range and rMSSD parameter range corresponding to Qi and Yin deficiency syndrome, Qi deficiency and blood stasis syndrome, heart and kidney Yang deficiency syndrome and Yang deficiency with water retention syndrome, respectively. The first mapping relationship is used as a label to label Qi and Yin deficiency syndrome, Qi deficiency and blood stasis syndrome, heart and kidney Yang deficiency syndrome and Yang deficiency and water retention syndrome, and the first label Qi and Yin deficiency syndrome, the first label Qi deficiency and blood stasis syndrome, the first label heart and kidney Yang deficiency syndrome and the first label Yang deficiency and water retention syndrome are obtained. The first label consists of the TCM syndrome text set of chronic heart failure, which includes the first label Qi and Yin deficiency syndrome, the first label Qi deficiency and blood stasis syndrome, the first label heart and kidney Yang deficiency syndrome, and the first label Yang deficiency and water retention syndrome.
4. The method for identifying TCM syndrome types in chronic heart failure according to claim 2, characterized in that, The second mapping relationship is used as a label to annotate the TCM syndrome type text set of chronic heart failure, resulting in the second-labeled TCM syndrome type text set of chronic heart failure, including: The second mapping relationship includes the LF parameter range, HF parameter range, and LF / HF parameter range corresponding to Qi and Yin deficiency syndrome, Qi deficiency and blood stasis syndrome, heart and kidney Yang deficiency syndrome, and Yang deficiency and water overflow syndrome, respectively. The second mapping relationship is used as a label to annotate the Qi and Yin deficiency syndrome, Qi deficiency and blood stasis syndrome, heart and kidney Yang deficiency syndrome and Yang deficiency and water retention syndrome, and the second label Qi and Yin deficiency syndrome, the second label Qi deficiency and blood stasis syndrome, the second label heart and kidney Yang deficiency syndrome and the second label Yang deficiency and water retention syndrome are obtained. The second label consists of the TCM syndrome text set of chronic heart failure, which includes the second label Qi and Yin deficiency syndrome, the second label Qi deficiency and blood stasis syndrome, the second label heart and kidney Yang deficiency syndrome, and the second label Yang deficiency and water retention syndrome.
5. The method for identifying TCM syndrome types in chronic heart failure according to claim 2, characterized in that, The first initial TCM syndrome identification model includes a first CNN module, an LSTM module, a first fully connected module, and a first Softmax module; The first initial TCM syndrome identification model was iteratively trained using a time-domain dataset and a first-label TCM syndrome type text set for chronic heart failure. After training, the first final TCM syndrome identification model was obtained, including: S11: Input the i-th time-domain data group in the time-domain dataset into the first CNN module, extract features from the SDNN and rMSSD parameters to obtain SDNN features and rMSSD features, and concatenate the SDNN features and rMSSD features to form the initial time-domain feature vector. ; S12: Initial time-domain feature vector The input LSTM module captures the dependencies in the features to obtain the temporal hidden state feature vector. ; transform the temporal hidden state feature vector The input is processed by the first fully connected module to perform feature dimensionality reduction and nonlinear transformation, resulting in the final time-domain feature vector. ; S13: The final time-domain feature vector Input the first Softmax module to perform probability prediction and obtain the first predicted probability distribution. Take the text of chronic heart failure TCM syndrome with the highest probability in the first predicted probability distribution as the first predicted chronic heart failure TCM syndrome text. S14: In the first-label chronic heart failure TCM syndrome type text set, obtain the first-label chronic heart failure TCM syndrome type text corresponding to the i-th time domain data group; calculate the first loss function value based on the probability of the first-label chronic heart failure TCM syndrome type text and the probability of the first predicted chronic heart failure TCM syndrome type text. S15: Backpropagate the first initial TCM syndrome identification model based on the first loss function value and adjust the parameters of the first initial TCM syndrome identification model. S16: Repeat steps S11-S15 until the first loss function value satisfies the first convergence condition, and obtain the first final TCM syndrome identification model.
6. The method for identifying TCM syndrome types in chronic heart failure according to claim 2, characterized in that, The second initial TCM syndrome identification model includes a second CNN module, a Transformer module, a second fully connected module, and a second Softmax module; The second initial TCM syndrome identification model is iteratively trained using a frequency domain dataset and a second-label TCM syndrome text set for chronic heart failure. After training, the second final TCM syndrome identification model is obtained, which includes: S21: Input the j-th frequency domain data group from the frequency domain data set into the second CNN module, extract features from the LF parameters, HF parameters, and LF / HF parameters to obtain LF features, HF features, and LF / HF features, and concatenate the LF features, HF features, and LF / HF features into an initial frequency domain feature vector. ; S22: Initial frequency domain feature vector The input is encoded and decoded by the Transformer module to obtain the frequency domain hidden state feature vector. The frequency domain hidden state feature vector The input is processed by the second fully connected module to perform feature dimensionality reduction and nonlinear transformation, resulting in the final frequency domain feature vector. ; S23: The final frequency domain feature vector Input the second Softmax module to perform probability prediction and obtain the second predicted probability distribution. The text of the chronic heart failure TCM syndrome with the highest probability in the second predicted probability distribution is taken as the second predicted chronic heart failure TCM syndrome text. S24: In the set of TCM syndrome type texts for chronic heart failure with second label, obtain the TCM syndrome type text for chronic heart failure with second label corresponding to the j-th frequency domain data group; calculate the second loss function value based on the probability of the TCM syndrome type text for chronic heart failure with second label and the probability of the second predicted TCM syndrome type text for chronic heart failure. S25: Backpropagate the second initial TCM syndrome identification model based on the second loss function value and adjust the parameters of the second initial TCM syndrome identification model. S26: Repeat steps S21-S25 until the value of the second loss function satisfies the second convergence condition, and obtain the second final TCM syndrome identification model.
7. The method for identifying TCM syndrome types in chronic heart failure according to claim 2, characterized in that, The first and second TCM syndrome probability sets are weighted and summed, and the maximum probability is selected to obtain the final TCM syndrome probability. Based on the final TCM syndrome probability, the final TCM syndrome text for chronic heart failure is obtained, including: The first set of probabilities for TCM syndrome types includes the probabilities of the first Qi and Yin deficiency syndrome, the first Qi deficiency and blood stasis syndrome, the first heart and kidney Yang deficiency syndrome, and the first Yang deficiency and water retention syndrome. The second set of probabilities for TCM syndrome types includes the probabilities of the second Qi and Yin deficiency syndrome, the second Qi deficiency and blood stasis syndrome, the second heart and kidney Yang deficiency syndrome, and the second Yang deficiency and water retention syndrome. The probability of the first Qi and Yin deficiency syndrome is obtained by weighted summing the probabilities of the second Qi and Yin deficiency syndrome; the probability of the first Qi deficiency and blood stasis syndrome is obtained by weighted summing the probabilities of the second Qi deficiency and blood stasis syndrome; the probability of the first heart and kidney Yang deficiency syndrome is obtained by weighted summing the probabilities of the second heart and kidney Yang deficiency syndrome; the probability of the first Yang deficiency and water retention syndrome is obtained by weighted summing the probabilities of the second Yang deficiency and water retention syndrome. The highest probability among the probabilities of Qi and Yin deficiency syndrome, Qi deficiency and blood stasis syndrome, heart and kidney Yang deficiency syndrome, and Yang deficiency and water retention syndrome is taken as the final TCM syndrome probability. The TCM syndrome text of chronic heart failure corresponding to the final TCM syndrome probability is taken as the final TCM syndrome text of chronic heart failure.
8. A device for identifying TCM syndrome types in chronic heart failure, used to implement the TCM syndrome type identification method for chronic heart failure as described in any one of claims 1 to 7, characterized in that, The device includes: The data acquisition module is used to acquire a heart rate variability dataset and a TCM syndrome type text set for chronic heart failure. The heart rate variability dataset includes a time domain dataset and a frequency domain dataset. The time-domain annotation module is used to obtain the first mapping relationship between the time-domain data set and the TCM syndrome type text set of chronic heart failure. The first mapping relationship is used as a label to annotate the TCM syndrome type text set of chronic heart failure, and the first label chronic heart failure TCM syndrome type text set is obtained. The frequency domain labeling module is used to obtain the second mapping relationship between the frequency domain data set and the TCM syndrome type text set of chronic heart failure. The second mapping relationship is used as a label to label the TCM syndrome type text set of chronic heart failure, and the second label chronic heart failure TCM syndrome type text set is obtained. The first initial TCM syndrome identification model training module is used to iteratively train the first initial TCM syndrome identification model using a time-domain data set and a first-label chronic heart failure TCM syndrome text set. After training, the first final TCM syndrome identification model is obtained. The second initial TCM syndrome identification model training module is used to iteratively train the second initial TCM syndrome identification model using a frequency domain data set and a second-label chronic heart failure TCM syndrome text set. After training, the second final TCM syndrome identification model is obtained. The TCM syndrome probability acquisition module is used to acquire the heart rate variability data set to be tested, divide the heart rate variability data set to be tested into a time domain data set to be tested and a frequency domain data set to be tested, input the time domain data set to be tested into a first final TCM syndrome identification model for TCM syndrome prediction, and obtain a first TCM syndrome probability set; input the frequency domain data set to be tested into a second final TCM syndrome identification model for TCM syndrome prediction, and obtain a second TCM syndrome probability set. The final TCM syndrome type text acquisition module for chronic heart failure is used to perform a weighted summation of the first TCM syndrome type probability set and the second TCM syndrome type probability set and select the maximum probability to obtain the final TCM syndrome type probability, and obtain the final TCM syndrome type text for chronic heart failure based on the final TCM syndrome type probability.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method for identifying TCM syndrome types of chronic heart failure as described in any one of claims 1 to 7.
10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for identifying TCM syndrome types of chronic heart failure as described in any one of claims 1 to 7.