AI multi-mode diagnosis method based on qi and blood theory of traditional Chinese medicine and chronic pain

Abstract

The invention provides an AI multi-modal diagnosis method based on the Qi and blood theory of traditional Chinese medicine and chronic pain, and the method comprises the following steps: 1, collecting retrospective research data and prospective research data, and carrying out the preprocessing and standardization of the data, and obtaining a multi-modal signal; step 2, carrying out normalization processing on the multi-modal signal, constructing a depth feature fusion model, and carrying out training and iterative optimization on the model; 3, performing model verification on the depth feature fusion model through an online incremental adaptive method; according to the system and the method, the traditional Chinese medicine diagnosis method and the modern medical technology are fused, and accurate capture and analysis of qi and blood characteristics of the chronic pain patient and generation of a treatment scheme are realized through highly automatic and intelligent means.

Description

Technical Field

[0001] The present invention relates to the technical field of traditional Chinese medicine syndrome differentiation for chronic pain, and particularly to an AI multi-modal diagnosis method based on the theory of qi and blood in traditional Chinese medicine and chronic pain. Background Art

[0002] Traditional Chinese medicine has a long history of understanding pain, and it is often classified and named according to the pain location, nature, zang-fu organs and meridians. As early as in the classic traditional Chinese medicine work "Huangdi Neijing", it is recorded that "pain is due to blockage and inability to pass through", "when the pulse is stagnant, there is blood deficiency, and blood deficiency leads to pain". Zhang Zihé pointed out in "Rumen Shiqin" that "all pains are caused by qi", indicating that pain is mainly based on the disorder of qi and blood in the body. Zhang Jingyue's "Doubts Record" and Li Dongyuan's "Medical Invention" respectively proposed that "pain due to deficiency of nourishment" and "pain due to obstruction" are the key pathogenesis of pain. To sum up, traditional Chinese medicine believes that the pathogenesis of chronic pain lies in "pain due to deficiency of nourishment" and "pain due to obstruction", and the disorder of qi and blood is the main cause of chronic pain. Chronic pain is characterized by a prolonged course that is difficult to cure, high disability and fatality rates, etc. Traditional Chinese medicine emphasizes the concept of wholism, reflects the pathogenesis syndrome of chronic pain patients through external physiological manifestations, and conducts syndrome differentiation and treatment, having unique advantages in the treatment of chronic pain. However, due to the strong subjectivity in the process of traditional Chinese medicine diagnosis and treatment and the lack of objective and quantitative diagnosis and treatment standards, the diagnostic results vary greatly, seriously affecting the treatment effect.

[0003] The traditional physiological and pathological parameters of qi and blood in traditional Chinese medicine for chronic pain are often subjectively diagnosed based on the clinical experience of doctors, mostly showing qualitative diagnosis, and their characteristic descriptions are somewhat fuzzy and difficult to quantify. Due to the personal experience of doctors, there are problems such as fragmentation and piecemeal collection of the physiological and pathological parameters manifested by qi and blood in traditional Chinese medicine for chronic pain. Summary of the Invention

[0004] The purpose of the present invention is to provide an AI multi-modal diagnosis method based on the theory of qi and blood in traditional Chinese medicine and chronic pain, which can solve the problems of strong subjectivity in the process of traditional Chinese medicine diagnosis and treatment and the lack of objective and quantitative diagnosis and treatment standards.

[0005] The purpose of the present invention is achieved through the following technical solutions:

[0006] An AI multi-modal diagnosis method based on the theory of qi and blood in traditional Chinese medicine and chronic pain, comprising the following steps:

[0007] Step 1: Collect retrospective research data and prospective research data, and preprocess and standardize the data to obtain multi-modal signals;

[0008] Step 2: Perform normalization processing on the multi-modal signals, construct a deep feature fusion model, and train and iteratively optimize the model;

[0009] Step 3: Verify the deep feature fusion model through an online incremental adaptive method;

[0010] Preferably, in Step 1, the retrospective study data includes traditional Chinese medicine classic monographs, online databases, and clinical electronic medical records; the prospective study data includes inspection, auscultation and olfaction, inquiry, and palpation; the specific operations for preprocessing and standardizing the data are: performing feature information processing and standardization processing on the data; the standardization processing is to form a standard thesaurus after standard classification and coding of the data, and update the data in the real-time updated standard thesaurus.

[0011] Preferably, in Step 2, perform imputation and frequency adjustment on the multi-modal signals to obtain data with frequency consistency; the specific operation of the imputation is to perform multiple imputations and mean filling on the data; the deep feature fusion model includes a recurrent neural network, a long short-term memory network, and a Transformer; the model training is specifically to input the preprocessed data, perform operations through the backpropagation algorithm, and continuously monitor the performance indicators of the deep feature fusion model; the iterative optimization is specifically to adjust the model structure and perform hyperparameter processing.

[0012] Preferably, in Step 3, the online incremental adaptive method includes data layer adaptation and feature layer adaptation; the data layer adaptation includes incremental data analysis and weighted difference term constraint; the feature layer adaptation includes feature detection and update and adaptive feature fusion; the model verification includes syndrome assessment verification and treatment plan recommendation; the passing of the syndrome assessment verification is achieved by designing a cross-sectional study and statistics; the treatment plan recommendation comes from the analysis of model rationality and the comparison between the model and expert suggestions.

[0013] The present invention integrates traditional Chinese medicine diagnosis methods and modern medical technologies, and through highly automated and intelligent means, realizes the accurate capture, analysis, and generation of treatment plans for the qi and blood characteristics of chronic pain patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic flowchart of an AI multi-modal diagnosis method based on traditional Chinese medicine qi and blood theory and chronic pain of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0015] The following describes the embodiments of the present disclosure in detail with reference to the drawings.

[0016] The following specific examples illustrate the implementation manners of the present disclosure. Those skilled in the art can easily understand the advantages and effects of the present disclosure from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of the embodiments. The present disclosure can also be implemented or applied through other different specific implementation manners. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that, without conflict, the following embodiments and the features in the embodiments can be combined with each other. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present disclosure without making creative efforts belong to the scope protected by the present disclosure.

[0017] Embodiment 1

[0018] As Figure 1 described, an AI multi-modal diagnosis method based on traditional Chinese medicine qi and blood theory and chronic pain includes the following steps:

[0019] Step 1: Collect retrospective study data and prospective study data, and preprocess and normalize the data to obtain multi-modal signals; through a self-developed four-diagnosis information collection device, combined with the classic theory of traditional Chinese medicine, comprehensively collect information on aspects such as observing, listening and smelling, inquiring, and pulse-taking of patients. Subsequently, preprocess and normalize these multi-source heterogeneous data to ensure data quality and lay a solid foundation for subsequent analysis.

[0020] Step 2: Normalize the multi-modal signals, construct a deep feature fusion model, and train and iteratively optimize the model; by adopting deep learning technology, construct an intelligent diagnosis and treatment model that can integrate multiple data types and learn the characteristics of chronic pain. During the model training process, the network architecture and hyperparameters will be continuously adjusted to optimize the model performance and ensure that it can accurately identify the pathological characteristics of chronic pain and perform effective syndrome classification.

[0021] Step 3: Verify the deep feature fusion model through an online incremental adaptive method; in order to improve the accuracy and reliability of the model in practical applications, an online incremental adaptive method is introduced, enabling the model to continuously adjust and optimize according to newly collected data and maintain its high sensitivity and specificity for the diagnosis of chronic pain. By comparing with the diagnosis results of clinical experts, the diagnostic accuracy of the model is verified, and iterative optimization is performed according to the feedback to continuously improve the model performance.

[0022] Further, in Step 1, the retrospective study data includes traditional Chinese medicine classic monographs, online databases, and clinical electronic medical records; the prospective study data includes inspection, auscultation and olfaction, inquiry, and palpation; the specific operations for preprocessing and standardizing the data are: performing feature information processing and standardization processing on the data; the standardization processing is to form a standard thesaurus after standard classification and coding of the data, and update the data in the real-time updated standard thesaurus.

[0023] Further, in Step 2, perform imputation and frequency adjustment on the multi-modal signals to obtain data with consistent frequency; the specific operation of the imputation is to perform multiple imputation and mean filling on the data; the deep feature fusion model includes a recurrent neural network, a long short-term memory network, and a Transformer; the model training is specifically to input the preprocessed data, perform operations through the backpropagation algorithm, and continuously monitor the performance indicators of the deep feature fusion model; the iterative optimization is specifically to adjust the model structure and perform hyperparameter processing.

[0024] Further, in Step 3, the online incremental adaptive method includes data layer adaptation and feature layer adaptation; the data layer adaptation includes incremental data analysis and weighted difference term constraint; the feature layer adaptation includes feature detection and update and adaptive feature fusion; the model verification includes syndrome evaluation verification and treatment plan recommendation; the passing of the syndrome evaluation verification is achieved by designing a cross-sectional study and statistics; the treatment plan recommendation comes from the analysis of model rationality and the comparison between the model and expert suggestions.

[0025] Online incremental adaptive method for the data layer model: There are general differences among patient individuals, which makes it difficult to construct a complete training data set, and it is possible for the old diagnostic model to produce biases when diagnosing new patients, that is, the "concept drift" problem of the diagnostic model. Therefore, the present invention intends to calculate the fitness between the incremental data and the old model, determine the severity of the change in data distribution, and adopt the method of adding a weighted difference term for fitness to constrain the online incremental adaptive process of the model.

[0026] Online incremental adaptive method for the feature layer model: In the process of constructing a machine learning model, traditional methods usually take minimizing the error between the output result and the expected result as the objective function, but it is difficult to achieve global parameter optimization for the problem of unequal lengths of global features caused by feature increments. Therefore, the present invention intends to approximate the global minimum through local minima under constraints to achieve the feature increment adaptation of the model.

[0027] The online incremental adaptive method for the data layer model and the online incremental adaptive method for the feature layer model solve the problem that the initial intelligent diagnosis model cannot adapt to the dynamically changing data distribution and feature dimensions.

[0028] In the present invention, deep learning techniques are comprehensively applied to process and analyze various types of medical data with the aim of deeply understanding the pathological state of patients. First, for image data, including images of the face, tongue, eyes, and skin, a convolutional neural network is used for feature extraction. The convolutional neural network is suitable for processing visual information due to its ability to capture the spatial hierarchical structure and details in images. For speech data, covering attributes such as the pitch, speech rate, and semantics of speech, as well as text data, such as patients' descriptions of pain, recurrent neural networks and their variants, the long short-term memory network and the gated recurrent unit, are used for processing. These models can effectively capture temporal dependencies and dynamic features due to their ability to process sequential data, thereby deeply understanding the complexity of speech and text data. For physiological signal data, such as skin temperature, humidity, and pulse, a fully connected layer network is used for feature extraction, aiming to extract key information from physiological signals with a relatively simple structure.

[0029] After the feature extraction of these different modality data, a Transformer model will be further adopted for feature fusion. The Transformer model, through its self-attention mechanism, has the advantage of being able to identify the internal connections between different modality data and achieve deep feature fusion. This fusion not only enhances the model's understanding of each data source, but also enables the model to capture cross-modal interactions and dependencies, generating a comprehensive and rich feature representation.

[0030] The present invention will focus on the correlation between the qi-blood characteristic information of chronic pain and the pathogenesis of traditional Chinese medicine, construct a diagnostic and adjuvant treatment model for intelligent multi-modal fusion of chronic pain, realize the intelligent and precise diagnosis and treatment of chronic pain in traditional Chinese medicine clinics, and improve the ability of traditional Chinese medicine clinics to treat pain. The present invention combines traditional Chinese medicine diagnosis methods with modern medical technologies, and through highly automated and intelligent means, realizes the precise capture, analysis of the qi-blood characteristics of chronic pain patients and the generation of treatment plans.

[0031] The above is only to illustrate the implementation mode of the present invention and is not intended to limit the present invention. For those skilled in the art, any modifications, equivalent replacements, improvements, etc. made without creative labor within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An AI multimodal diagnostic method based on TCM Qi and blood theory and chronic pain, characterized in that: The following steps are involved: Step 1: Collect retrospective and prospective study data, preprocess and normalize the data to obtain multimodal signals; Step 2: Normalize the multimodal signals, build a deep feature fusion model, and train and iteratively optimize the model; Step 3: Verify the deep feature fusion model through the online incremental adaptive method.

2. The AI ​​multimodal diagnosis method based on TCM Qi and blood theory and chronic pain as claimed in claim 1, characterized in that: In step one, the retrospective research data include classic monographs of traditional Chinese medicine, online databases and clinical electronic medical records; the prospective research data include inspection, auscultation, questioning and palpation; the specific operations of preprocessing and normalizing the data are: processing the feature information and standardizing the data; the standardization process is to classify and encode the data standards to form a standard vocabulary, and update the data in the standard vocabulary in real time.

3. The AI ​​multimodal diagnosis method based on TCM Qi and blood theory and chronic pain as claimed in claim 1, characterized in that: In step two, the multimodal signal is interpolated and frequency adjusted to obtain frequency-consistent data; the specific operation of the interpolation is multiple interpolation and mean padding of the data; the deep feature fusion model includes a recurrent neural network, a long short-term memory network and a Transformer; the model training specifically inputs preprocessed data, performs calculations through a back propagation algorithm, and continuously monitors the performance indicators of the deep feature fusion model; the iterative optimization specifically adjusts the model structure and performs hyperparameter processing.

4. The AI ​​multimodal diagnosis method based on TCM Qi and blood theory and chronic pain as claimed in claim 1, characterized in that: In step three, the online incremental adaptation method includes data layer adaptation and feature layer adaptation; The data layer adaptation includes incremental data analysis and weighted difference term constraints; The feature layer adaptation includes feature detection and updating and adaptive feature fusion; the model validation includes syndrome evaluation validation and treatment plan recommendation; the syndrome evaluation validation is achieved by designing a cross-sectional study and statistics; the treatment plan recommendation comes from the model rationality analysis and the comparison between the model and expert advice.

Images