Depression assessment system based on medical data processing

Abstract

This paper proposes a depression assessment system based on medical data processing. The data fusion module extracts features from medical text, voice data, and physiological electrical signal data through a multi-branch neural network, and adopts a bidirectional cross-attention mechanism for dynamic interaction. The assessment and optimization module assigns confidence scores to feature sets to determine whether the data is sensitive information. Sensitive data is processed by an authoritative knowledge base, while non-sensitive data is processed by a general knowledge base. The multi-branch neural network parameters are updated simultaneously through an adaptive learning rate. The state assessment module uses an MM-RNN model to calculate the depression state score and determine the depression level. By introducing a sensitive information determination mechanism based on confidence scores, the present invention intelligently assesses data quality and dynamically adjusts the data processing path to ensure the effective protection of sensitive data, improves the accuracy and reliability of the assessment results, and enhances the adaptability and security of the system in processing various medical data.

Description

Technical Field

[0001] The present invention relates to the technical field of depression assessment, and in particular to a depression assessment system based on medical data processing. Background Art

[0002] Depression is a common mental illness that affects a large number of people worldwide. Its symptoms usually include low mood, loss of interest, sleep disorders, etc. In severe cases, it can affect the patient's normal life and work. Traditional depression assessment methods mainly rely on clinical diagnosis and standardized questionnaires, such as the Hamilton Depression Rating Scale and the Beck Depression Inventory, to assess patients' depressive symptoms through self-reports and doctor observations. With the advancement of computer technology and big data analysis, depression assessment systems have been proposed to assist in the diagnosis and assessment of depression, thereby providing more accurate and objective assessment results.

[0003] Existing depression assessment systems mostly rely on static data analysis, such as a single scale score or single-modality physiological signal data analysis. This approach ignores the dynamic relationship and interaction between different data modalities, resulting in incomplete and inaccurate assessment results. At the same time, existing depression assessment systems have a relatively simple approach to data quality and lack in-depth exploration and weighing of the mutual influence between different types of data. As a result, when faced with complex and changeable clinical data, the system's assessment results are insufficiently accurate and robust. Summary of the Invention

[0004] In order to solve the technical problems mentioned in the current background technology, the purpose of the present invention is to provide a depression assessment system based on medical data processing.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] The depression assessment system based on medical data processing includes:

[0007] M1, data fusion module, extracts modal features from medical text data, voice data, and physiological electrical signal data through a multi-branch neural network, and uses a bidirectional cross-attention mechanism to dynamically interact with the modal features to generate a structured feature set;

[0008] M2, an evaluation and optimization module, performs a confidence score on the structured feature set, the confidence score is weightedly calculated by the medical text feature stability index, the speech clarity factor, and the physiological electrical signal factor, and the confidence score is compared with the confidence score threshold to determine whether it is sensitive information. The corresponding search path for sensitive information is the authoritative knowledge base, and the corresponding search path for non-sensitive information is the general knowledge base, and the optimal medical data reference set is output;

[0009] Based on the optimal medical data reference set, an adaptive learning rate strategy is used to iteratively update the parameters of the multi-branch neural network;

[0010] M3, a state assessment module, dynamically weights the reference feature vectors within the optimal medical data reference set, calculates the depression state score based on the processed reference feature vectors using the MM-RNN model, generates a dynamic depression threshold for depression state assessment according to a dynamic threshold function; and determines the depression state level based on the dynamic threshold.

[0011] Furthermore, the medical text data mainly comes from the patient's medical records and doctor's diagnosis reports; the voice data mainly comes from the patient's voice symptom descriptions and related voice records; the physiological electrical signal data mainly comes from the patient's electrocardiogram and electroencephalogram;

[0012] The multi-branch neural network includes a BERT model, a self-supervised speech model and a CNN-LSTM hybrid model;

[0013] The process of extracting the modal features is as follows: the medical text data is extracted through the BERT model to obtain the medical text modal feature set. The speech data is extracted into a speech modality feature set through a self-supervised speech model The physiological electrical signal data is extracted through the CNN-LSTM hybrid model to obtain the physiological electrical signal modal feature set .

[0014] Furthermore, the dynamic interaction process is as follows:

[0015] Inputting the medical text modality feature set, the speech modality feature set, and the physiological electrical signal modality feature set into a bidirectional cross-attention mechanism to obtain an attention matrix, and then dynamically weighting the medical text modality feature set, the speech modality feature set, and the physiological electrical signal modality feature set using the attention matrix;

[0016] Dynamically interacting medical text modality feature sets through the bidirectional cross-attention mechanism , speech modality feature set and physiological electrical signal modality feature set Features between;

[0017] After the bidirectional cross attention mechanism is calculated, the obtained attention matrix is ​​used to analyze the medical text modality feature set. , speech modality feature set and physiological electrical signal modality feature set Perform dynamic weighted fusion, the calculation formula is:

[0018] ;

[0019] ;

[0020] ;

[0021] in, is the modal feature set of medical text after dynamic interaction , is the medical text modality feature set Speech modality feature set The fused features; is the modal feature set of physiological electrical signals after dynamic interaction , is the medical text modality feature set and physiological electrical signal modality feature set The fused features; is the speech modality feature set after dynamic interaction , is the speech modality feature set and physiological electrical signal modality feature set The fused features; is the medical text modality feature set and speech modality feature set The attention matrix between them; is the medical text modality feature set and physiological electrical signal modality feature set The attention matrix between them; is the speech modality feature set and physiological electrical signal modality feature set The attention matrix between

[0022] The structured feature set is denoted as F, and .

[0023] Furthermore, the confidence score calculation step includes calculating the medical text feature stability index, calculating the speech clarity factor, and calculating the physiological electrical signal clarity factor;

[0024] The medical text feature stability index The calculation process is to calculate the medical text modality feature set The mean of the medical text modality feature set is calculated The standard deviation and the final calculation of the medical text feature stability index ;

[0025] The speech intelligibility factor The calculation process is to calculate the power of the speech signal, calculate the power of the noise in the speech signal and finally calculate the speech clarity factor ;

[0026] The physiological electrical signal clarity factor The calculation process is to calculate the power of the physiological electrical signal, calculate the power of the noise in the physiological electrical signal and finally calculate the clarity factor of the physiological electrical signal. , and finally the physiological electrical signal clarity factor is obtained ;

[0027] Based on the medical text feature stability index , Speech Intelligibility Factor and physiological electrical signal clarity factor Calculate the confidence score , the calculation formula is:

[0028] ;

[0029] in, 、 and is the weight coefficient, and ; Corresponding medical text feature stability index The weight coefficient of Corresponding speech clarity factor The weight coefficient of Corresponding physiological electrical signal clarity factor The weight coefficient of

[0030] The confidence score threshold is set to 0.8; when When , it is judged as sensitive information, and the corresponding search path is the authoritative knowledge base; when , it is judged as non-sensitive information, and the corresponding search path is the general knowledge base; finally, the optimal medical data reference set is output according to the search results of different paths.

[0031] Furthermore, the optimal medical data reference set is recorded as ,and ,in, For the The reference feature vector of medical text data, speech data and physiological electrical signal data after linear fusion;

[0032] The reference feature vectors within the optimal medical data reference set are dynamically weighted, and the calculation formula is:

[0033] ;

[0034] in, is The moment The weight coefficient of the reference eigenvector; is the number of reference eigenvectors; It is with The time point at which the reference eigenvector is most relevant; is a smoothing parameter used to control the sensitivity of weighting;

[0035] During the evaluation process, the weighted dynamic feature set needs to be calculated first through dynamic weighting , the calculation formula is:

[0036] ;

[0037] Based on dynamic feature set , using the MM-RNN model, after dynamically learning the temporal dependency of the reference feature vector, the depression state score is calculated:

[0038] ;

[0039] in, is the output depression state score, ranging from [0,1], indicating the patient's current depression state; is the weight matrix of the output layer; ; is in the time step The hidden state when is the activation function;

[0040] The generating function of the dynamic depression threshold is as follows:

[0041] ;

[0042] B ;

[0043] in, and All are adjusted dynamic depression thresholds; is the initial set depression threshold; is the amplitude of variation factor; is the attenuation factor that controls the rate of state change; is a Gaussian noise term used to simulate natural fluctuations in psychological state; is the offset constant;

[0044] Determine the level of depression based on the dynamic depression threshold:

[0045] ;

[0046] in, Mild depression; Moderate depression; For severe depression.

[0047] Furthermore, the features of the optimal medical data reference set are used as input data to calculate the loss function , the loss function Depends on the parameters , in each iteration, the loss function is calculated Current network parameters Gradient , while the learning rate Dynamic adjustment is performed based on the historical information of the gradient, and finally based on the gradient and adaptive learning rate Update the multi-branch neural network parameters .

[0048] Compared with the prior art, the advantages of the present invention are:

[0049] 1. The present invention combines medical text, voice data and physiological electrical signal data, and adopts a multi-branch neural network and a two-way cross-attention mechanism to fully explore the deep relationship between various data modalities, so that data of different modalities can be weighted according to their own importance during the dynamic interaction process, avoiding the one-sided evaluation caused by a single data source in traditional methods. Especially in the early diagnosis of depression, the patient's language expression and physiological signals often do not have obvious symptoms. By dynamically integrating multiple data, this system can achieve a more comprehensive and accurate evaluation, thereby improving the diagnostic accuracy of depression.

[0050] 2. The present invention introduces a dynamic weighting mechanism based on confidence assessment, and dynamically evaluates the reliability of various types of data by calculating the medical text stability index, speech clarity factor and physiological electrical signal factor. This innovative mechanism can adaptively adjust to data of different quality to ensure that data noise or missing data will not have an excessive impact on the evaluation results. In practical applications, the system can optimize the weight distribution of each data modality in real time according to the quality of each data input, thereby improving the stability and reliability of the evaluation results for different patients and different clinical environments.

[0051] 3. The present invention significantly improves the system's ability to process medical data of different qualities through a sensitive information determination mechanism based on confidence scoring. The mechanism can intelligently identify the sensitivity of the data and automatically select the appropriate data path according to the confidence of the data, ensuring that sensitive data is strictly protected and improving the efficiency of non-sensitive data processing. Through dynamic evaluation of medical data quality, the system can effectively avoid the risk of erroneous evaluation or data leakage caused by differences in data quality, thereby ensuring privacy protection while improving the accuracy and reliability of the evaluation results. In addition, the mechanism enhances the flexibility and adaptability of the system, so that when faced with complex clinical data, it can more efficiently process multiple data sources, thereby improving the overall performance and clinical application value of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0053] Figure 1 Schematic diagram of the system workflow of the present invention;

[0054] Figure 2 Schematic diagram of the structured feature set calculation process of the present invention;

[0055] Figure 3 Schematic diagram of the calculation process of the optimal medical data reference set of the present invention. DETAILED DESCRIPTION

[0056] To make the purpose, technical solutions, and advantages of the embodiments of this application more clear, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0057] To achieve the above objectives, the present invention is implemented through the following technical solutions: the present invention provides a depression assessment system based on medical data processing, such as Figure 1-Figure 3 As shown, the system includes:

[0058] M1, the data fusion module, extracts modal features from medical text data, voice data, and physiological electrical signal data through a multi-branch neural network, and uses a bidirectional cross-attention mechanism to dynamically interact with the modal features to generate a structured feature set.

[0059] The medical text data mainly comes from the patient's medical records and doctor's diagnosis reports; the voice data mainly comes from the patient's voice symptom description and related voice records; the physiological electrical signal data mainly comes from the patient's electrocardiogram and electroencephalogram;

[0060] The multi-branch neural network includes a BERT model, a self-supervised speech model and a CNN-LSTM hybrid model;

[0061] The process of extracting the modal features is as follows: the medical text data is extracted through the BERT model to obtain the medical text modal feature set. The speech data is extracted into a speech modality feature set through a self-supervised speech model The physiological electrical signal data is extracted through the CNN-LSTM hybrid model to obtain the physiological electrical signal modal feature set ;

[0062] The dynamic interaction process is as follows:

[0063] Inputting the medical text modality feature set, the speech modality feature set, and the physiological electrical signal modality feature set into a bidirectional cross-attention mechanism to obtain an attention matrix, and then using the attention matrix to dynamically weighted fuse the medical text modality feature set, the speech modality feature set, and the physiological electrical signal modality feature set;

[0064] Dynamically interacting medical text modality feature sets through the bidirectional cross-attention mechanism , speech modality feature set and physiological electrical signal modality feature set The features between are calculated as follows:

[0065] ;

[0066] ;

[0067] ;

[0068] in, is the medical text modality feature set and speech modality feature set The attention matrix between them; is the medical text modality feature set and physiological electrical signal modality feature set The attention matrix between them; is the speech modality feature set and physiological electrical signal modality feature set The attention matrix between them; 、 and To calculate the attention weight matrix, it is used to calculate the similarity between different modality pairs; is the dimension of the feature vector, used to normalize the attention score.

[0069] After the bidirectional cross attention mechanism is calculated, the obtained attention matrix is ​​used to analyze the medical text modality feature set. , speech modality feature set and physiological electrical signal modality feature set Perform dynamic weighted fusion, the calculation formula is:

[0070] ;

[0071] ;

[0072] ;

[0073] in, is the modal feature set of medical text after dynamic interaction , is the medical text modality feature set Speech modality feature set The fused features; is the modal feature set of physiological electrical signals after dynamic interaction , is the medical text modality feature set and physiological electrical signal modality feature set The fused features; is the speech modality feature set after dynamic interaction , is the speech modality feature set and physiological electrical signal modality feature set The fused features;

[0074] The structured feature set is denoted as F, and .

[0075] M2, an evaluation and optimization module, performs a confidence score on the structured feature set, the confidence score is dynamically weighted by the medical text feature stability index, the speech clarity factor, and the physiological electrical signal factor, and the confidence score is compared with the confidence score threshold to determine whether it is sensitive information. The search path corresponding to sensitive information is the authoritative knowledge base, and the search path corresponding to non-sensitive information is the general knowledge base, and the optimal medical data reference set is output;

[0076] The confidence score calculation step includes calculating the medical text feature stability index, calculating the speech clarity factor, and calculating the physiological electrical signal clarity factor;

[0077] The medical text feature stability index The calculation process is to calculate the medical text modality feature set The mean of the medical text modality feature set is calculated The standard deviation and the final calculation of the medical text feature stability index , the calculation formula is:

[0078] ;

[0079] in, is the medical text modality feature set The mean of is the number of modal features of medical text; For the modal features of medical texts;

[0080] ;

[0081] in, is the medical text modality feature set The standard deviation of

[0082] ;

[0083] in, is the medical text feature stability index.

[0084] The speech intelligibility factor The calculation process is to calculate the power of the speech signal, calculate the power of the noise in the speech signal and finally calculate the speech clarity factor. The calculation formula is:

[0085] ;

[0086] in, is the power of the speech signal; is the duration of the speech signal; The speech signal at time The value on , that is, the amplitude of the speech signal;

[0087] ;

[0088] in, is the power of the noise in the speech signal, indicating the intensity of the background noise; is the duration of the noise signal; Noise signal in time The value on

[0089] ;

[0090] in, is the signal-to-noise ratio of the speech signal;

[0091] ;

[0092] in, is the speech intelligibility factor; It is a coefficient related to the clarity of the speech signal and is used to adjust the influence of the factor.

[0093] The physiological electrical signal clarity factor The calculation process is to calculate the power of the physiological electrical signal, calculate the power of the noise in the physiological electrical signal and finally calculate the clarity factor of the physiological electrical signal. The calculation process is the same as the speech clarity factor The calculation process finally results in the physiological electrical signal clarity factor ;

[0094] Based on the medical text feature stability index , Speech Intelligibility Factor and physiological electrical signal clarity factor Calculate the confidence score , the calculation formula is:

[0095] ;

[0096] in, 、 and is the weight coefficient, and ; Corresponding medical text feature stability index The weight coefficient of Corresponding speech clarity factor The weight coefficient of Corresponding physiological electrical signal clarity factor The weight coefficient of

[0097] In this embodiment, , , ;

[0098] The confidence score threshold is set to 0.8; when When , it is judged as sensitive information, and the corresponding search path is the authoritative knowledge base; when , it is judged as non-sensitive information, and the corresponding search path is the general knowledge base; finally, the optimal medical data reference set is output according to the search results of different paths.

[0099] The features of the optimal medical data reference set are used as input data to calculate the loss function , the loss function Depends on the parameters , the calculation formula is:

[0100] ;

[0101] in, is the number of characteristic samples; is the true value, that is, the actually observed data; is the predicted value, that is, the predicted output of the multi-branch neural network under the current parameters; For the The parameters of the multi-branch neural network for the iteration;

[0102] In each iteration, the loss function is calculated Current network parameters Gradient , the calculation formula is:

[0103] ;

[0104] in, For the The gradient of the iteration represents the loss function About the parameters of multi-branch neural networks The partial derivative of is the partial derivative;

[0105] In each iteration, the learning rate Dynamic adjustment is performed based on the historical information of the gradient to adapt to the data and gradient changes during network training. The calculation formula is:

[0106] ;

[0107] in, For the The weighted average of the squares of the gradients in iterations; No. The weighted average of the squares of the gradients in iterations; is the decay rate, and ;

[0108] ;

[0109] in, is the adaptive learning rate; is the initial learning rate; is a constant, and ;

[0110] In this embodiment, ;

[0111] Gradient-based and adaptive learning rate Update multi-branch neural network parameters , the calculation formula is:

[0112] ;

[0113] in, For the Multi-branch neural network parameters after iterations; For the The parameters of the multi-branch neural network at the iteration.

[0114] M3, a state assessment module, which performs weighted processing on the reference feature vectors within the optimal medical data reference set through dynamic weighting, calculates a depression state score based on the processed reference feature vectors using the MM-RNN model, generates a dynamic depression threshold for depression state assessment based on a dynamic threshold function, and determines the depression state level based on the dynamic threshold;

[0115] The optimal medical data reference set is denoted as ,and ,in, For the The reference feature vector of the weighted fusion of medical text data, speech data and physiological electrical signal data after linear fusion;

[0116] The reference feature vectors within the optimal medical data reference set are weighted by feature weighting, and the calculation formula is:

[0117] ;

[0118] in, is The moment The weight coefficient of the reference eigenvector; is the number of reference eigenvectors; It is with The time point at which the reference eigenvector is most relevant; is a smoothing parameter used to control the sensitivity of weighting;

[0119] In this embodiment, ;

[0120] During the evaluation process, the weighted dynamic feature set needs to be calculated first through dynamic weighting , the calculation formula is:

[0121] ;

[0122] Based on dynamic feature set , using the MM-RNN model, the temporal dependency of the reference feature vector is dynamically learned, and the calculation formula is:

[0123] ;

[0124] in, is in the time step The hidden state at time , represents the reference feature vector representation of the current time step; is an activation function used to introduce nonlinear relationships; is the weight matrix of the hidden layer, which is used to connect the hidden state of the previous moment and the input of the current moment; is the hidden state of the previous time step; is the bias term of the hidden layer.

[0125] Finally, the depression state score was calculated:

[0126] ;

[0127] in, is the output depression state score, ranging from [0,1], indicating the patient's current depression state; is the weight matrix of the output layer; is the bias term of the output layer;

[0128] The generating function of the dynamic depression threshold is as follows:

[0129] ;

[0130] B ;

[0131] in, and All are adjusted dynamic depression thresholds; is the initial depression threshold, and ; is the amplitude factor, and The value range of 【0.1,0.3】; A decay factor that controls the rate of state change; is a Gaussian noise term used to simulate natural fluctuations in psychological state; is the offset constant, and ;

[0132] Determine the level of depression based on the dynamic depression threshold:

[0133] ;

[0134] in, Mild depression; Moderate depression; For severe depression.

[0135] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0136] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A depression assessment system based on medical data processing, characterized in that: include: M1, data fusion module, extracts modal features from medical text data, voice data, and physiological electrical signal data through a multi-branch neural network, and uses a bidirectional cross-attention mechanism to dynamically interact with the modal features to generate a structured feature set; M2, an evaluation and optimization module, performs a confidence score on the structured feature set, the confidence score is weightedly calculated by the medical text feature stability index, the speech clarity factor, and the physiological electrical signal factor, and the confidence score is compared with the confidence score threshold to determine whether it is sensitive information. The corresponding search path for sensitive information is the authoritative knowledge base, and the corresponding search path for non-sensitive information is the general knowledge base, and the optimal medical data reference set is output; The confidence score threshold is set to 0.8; when When , it is judged as sensitive information, and the corresponding search path is the authoritative knowledge base; when When , it is judged as non-sensitive information, and the corresponding search path is the general knowledge base; finally, the optimal medical data reference set is output according to the search results of different paths; Based on the optimal medical data reference set, an adaptive learning rate strategy is used to iteratively update the parameters of the multi-branch neural network; M3, a state assessment module, dynamically weights the reference feature vectors within the optimal medical data reference set, calculates a depression state score based on the processed reference feature vectors using the MM-RNN model, generates a dynamic depression threshold for depression state assessment based on a dynamic threshold function, and determines the depression state level based on the dynamic threshold; The medical text data is derived from the patient's medical records and doctor's diagnosis report; the voice data is derived from the patient's voice symptom description and related voice records; the physiological electrical signal data is derived from the patient's electrocardiogram and electroencephalogram; The multi-branch neural network includes a BERT model, a self-supervised speech model and a CNN-LSTM hybrid model; The process of extracting the modal features is as follows: the medical text data is extracted through the BERT model to obtain the medical text modal feature set. The speech data is extracted into a speech modality feature set through a self-supervised speech model The physiological electrical signal data is extracted through the CNN-LSTM hybrid model to obtain the physiological electrical signal modal feature set .

2. The depression assessment system based on medical data processing according to claim 1, characterized in that: The dynamic interaction process is as follows: Inputting the medical text modality feature set, the speech modality feature set, and the physiological electrical signal modality feature set into a bidirectional cross-attention mechanism to obtain an attention matrix, and then dynamically weighting the medical text modality feature set, the speech modality feature set, and the physiological electrical signal modality feature set using the attention matrix; Dynamically interacting medical text modality feature sets through the bidirectional cross-attention mechanism Speech modality feature set and physiological electrical signal modality feature set Features between; After the bidirectional cross attention mechanism is calculated, the obtained attention matrix is ​​used to analyze the medical text modality feature set. , speech modality feature set and physiological electrical signal modality feature set Perform dynamic weighted fusion, the calculation formula is: ； ； ； in, is the modal feature set of medical text after dynamic interaction , is the medical text modality feature set Speech modality feature set The fused features; is the modal feature set of physiological electrical signals after dynamic interaction , is the medical text modality feature set and physiological electrical signal modality feature set The fused features; is the speech modality feature set after dynamic interaction , is the speech modality feature set and physiological electrical signal modality feature set The fused features; is the medical text modality feature set and speech modality feature set The attention matrix between them; is the medical text modality feature set and physiological electrical signal modality feature set The attention matrix between them; is the speech modality feature set and physiological electrical signal modality feature set The attention matrix between them; The structured feature set is denoted as F, and .

3. The depression assessment system based on medical data processing according to claim 1, characterized in that: The confidence score calculation step includes calculating the medical text feature stability index, calculating the speech clarity factor, and calculating the physiological electrical signal clarity factor; The medical text feature stability index The calculation process is to calculate the medical text modality feature set The mean of the medical text modality feature set is calculated The standard deviation and the final calculation of the medical text feature stability index ; The speech intelligibility factor The calculation process is to calculate the power of the speech signal, calculate the power of the noise in the speech signal and finally calculate the speech clarity factor ; The physiological electrical signal factor The calculation process is to calculate the power of the physiological electrical signal, calculate the power of the noise in the physiological electrical signal and finally calculate the physiological electrical signal factor , and finally the physiological electrical signal clarity factor is obtained ; Based on the medical text feature stability index , Speech Intelligibility Factor and physiological electrical signal clarity factor Calculate the confidence score , the calculation formula is: ； in, 、 and is the weight coefficient, and ; Corresponding medical text feature stability index The weight coefficient of Corresponding speech clarity factor The weight coefficient of Corresponding physiological electrical signal clarity factor The weight coefficient of .

4. The depression assessment system based on medical data processing according to claim 1, characterized in that: The optimal medical data reference set is denoted as ,and ,in, For the The reference feature vector of medical text data, speech data and physiological electrical signal data after linear fusion; The reference feature vectors within the optimal medical data reference set are dynamically weighted, and the calculation formula is: ； in, is The moment The weight coefficient of the reference eigenvector; is the number of reference eigenvectors; It is with The time point at which the reference eigenvector is most relevant; is a smoothing parameter used to control the sensitivity of weighting; During the evaluation process, the weighted dynamic feature set needs to be calculated first through dynamic weighting , the calculation formula is: ； Based on dynamic feature set , using the MM-RNN model, after dynamically learning the temporal dependency of the reference feature vector, the depression state score is calculated: ； in, is the output depression state score, ranging from [0,1], indicating the patient's current depression state; is the weight matrix of the output layer; is the bias term of the output layer; The generating function of the dynamic depression threshold is as follows: ； in, and All are adjusted dynamic depression thresholds; is the initial set depression threshold; is the change amplitude factor; the attenuation factor that controls the rate of state change; is a Gaussian noise term used to simulate natural fluctuations in psychological state; is the offset constant; Determine the level of depression based on the dynamic depression threshold: ； in, Mild depression; Moderate depression; For severe depression.

5. The depression assessment system based on medical data processing according to claim 1, characterized in that: The features of the optimal medical data reference set are used as input data to calculate the loss function , the loss function Depends on the parameters , in each iteration, the loss function is calculated Current network parameters Gradient , while the learning rate Dynamic adjustment is performed based on the historical information of the gradient, and finally based on the gradient and adaptive learning rate Update the multi-branch neural network parameters .

Images