Auxiliary diagnosis processing method based on artificial intelligence and human body image acupuncture points

Abstract

The invention relates to the technical field of artificial data analysis, in particular to an auxiliary diagnosis processing method based on artificial intelligence and human body image acupuncture points, which comprises the following steps: S1, receiving a temperature matrix generated by a temperature sensor array and a dictionary array containing acupuncture point coordinates and name information; s2, setting a radius range for each acupoint, obtaining temperature values of all points in the range, calculating an average temperature, and generating corresponding regional temperature data; s3, accurately associating the acupoint names with the corresponding temperature features to form a structured acupoint-coordinate-temperature relation dictionary model; s4, generating an accurate meridian temperature visualization map according to the temperature characteristics; s5, performing auxiliary diagnosis on the acupoint coordinate temperature dictionary based on a Transform network, and generating an artificial intelligence model diagnosis result; s6, docking the generated meridian and collateral temperature map with the diagnosis demand of a doctor through a cloud diagnosis system; the system integrates traditional Chinese medicine meridian theory, data analysis and artificial intelligence technology, and assists doctors in diagnosis.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial data analysis, and in particular, to an auxiliary diagnosis and processing method based on artificial intelligence and human body image acupoints. Background Art

[0002] According to traditional Chinese medicine meridian theory, the physiological state of human acupoints is closely related to overall health. By observing and diagnosing acupoints, the health status of patients can be effectively understood. However, in traditional Chinese medicine diagnosis and treatment, the analysis of meridians and acupoints usually relies on doctors' experience and lacks objective and quantitative data support. In particular, the research on acupoint temperature is rare.

[0003] Currently, there is no system that can efficiently combine traditional Chinese medicine meridian theory, modern data processing, and artificial intelligence technology to intuitively display the temperature state of human acupoints, the effective auxiliary diagnosis results, and the auxiliary diagnosis results made by artificial intelligence, and assist doctors in diagnosis analysis and treatment decision-making. Summary of the Invention

[0004] The purpose of the present invention is to overcome the deficiencies of the prior art, solve the technical blank of the lack of a complete acupoint temperature curve in the traditional meridian system and the problem of low intelligence in data analysis by artificial intelligence, and propose an auxiliary diagnosis and processing method based on artificial intelligence and human body image acupoints.

[0005] The purpose of the present invention is achieved by the following measures: An auxiliary diagnosis and processing method based on artificial intelligence and human body image acupoints, comprising the following steps:

[0006] S1. Receive the temperature matrix generated by the temperature sensor array and the dictionary array containing acupoint coordinate and name information;

[0007] S2. For each acupoint, set a radius range with it as the center, obtain the temperature values of all points within the range and calculate the average temperature to generate corresponding regional temperature data;

[0008] S3. Precisely associate the acupoint name with the corresponding temperature feature to form a structured acupoint - coordinate - temperature relationship dictionary model;

[0009] S4. Based on traditional Chinese medicine meridian theory, arrange the acupoints in different meridian systems in a specific order, and generate a precise meridian temperature visualization map according to their temperature features;

[0010] S5. Based on the Transformer network, perform auxiliary diagnosis on the acupoint coordinate temperature dictionary to generate the diagnosis result of the artificial intelligence model;

[0011] S6. Connect the generated meridian temperature map with the doctor's diagnosis requirements through the cloud diagnosis system to assist the doctor in conducting health analysis and treatment recommendations based on the temperature chart and the auxiliary diagnosis results obtained from the artificial intelligence algorithm.

[0012] Preferably, the temperature matrix in step S1 is a multi-dimensional matrix generated by a temperature sensor array. The temperature matrix records the temperature values of each point in the area, and each matrix element corresponds to the temperature value of a spatial coordinate point. The dictionary stores the coordinate and name information of multiple acupoints.

[0013] Preferably, the calculation method of step S2 is as follows:

[0014]

[0015] Where L is the distance between the matrix element and the acupoint, Xi, Yi are the horizontal and vertical coordinates of different acupoints, Xn, Yn are the horizontal and vertical coordinates of a certain matrix element in the temperature matrix, i is the number of acupoints, and n is the number of matrix elements;

[0016] If L ≤ R, it is determined that the matrix element is a valid element for calculating the temperature value of a certain acupoint. If L > R, it is determined that the matrix element is an invalid element for calculating the temperature value of a certain acupoint;

[0017]

[0018] Where Ti is the regional temperature average value of a certain acupoint, Tx is the temperature value of each valid matrix element, and N is the total number of valid elements.

[0019] Preferably, step S3 includes:

[0020] For each acupoint name and its corresponding coordinates (Xi, Yi) in the new dictionary, generate a new mapping relationship through the temperature average value calculated in step S2, where the acupoint name is directly associated with its regional temperature value;

[0021] The newly generated dictionary structure is: acupoint name: coordinates Xi, Yi, temperature Ti, and the temperature value Ti can be stored with different precisions according to the actual needs of different parts. Support decimal precision control (for example, retain two decimal places) for refined diagnosis requirements.

[0022] Preferably, step S4 includes:

[0023] Classify all known meridians, determine the specific acupoints included in each meridian, and sort them according to the order of the acupoints and their temperature values in each meridian to ensure that the temperature data matches the physiological characteristics of the meridians;

[0024] Match the temperature values of each acupoint in the meridian with its position relationship on the human body to ensure that the temperature value of each acupoint can accurately reflect the physiological state of its location, and generate corresponding temperature layers in sequence according to the temperature values;

[0025] For the temperature map of each meridian, adopt a suitable color mapping method and use heat map and gradient color bar visualization methods to ensure that the areas corresponding to different temperature values can be clearly distinguishable; help doctors intuitively understand the temperature status of different acupoints and their possible health problems;

[0026] According to the diagnostic requirements and actual situation, adjust the display method of the temperature map of each meridian, including the refined processing of temperature data and the removal of outliers. Further improve the accuracy and practicality of the chart.

[0027] Preferably, in step S5, first, organize and format the input acupoint coordinates and temperature data, organize the data into a dictionary format, normalize the coordinate and temperature data, scale them to the interval [0,1] respectively, and finally, embed the normalized data into a high-dimensional space to prepare for the input of the Transformer model;

[0028] Dictionary formatting, for example: {"acupoint 1":(coordinate x1, y1, temperature temp1), "acupoint 2":(coordinate x2, y2, temperature temp2),...}.

[0029] Design the encoder part of the Transformer, use the multi-head self-attention mechanism (Multi-Head Self-Attention) to extract the global dependency relationship between acupoint coordinates and temperature. At the same time, introduce positional encoding (PositionalEncoding) to retain the spatial information of acupoints. In each encoder layer, add a feed-forward network and layer normalization operations; to enhance the expression ability and stability of the model.

[0030] Preferably, in order to train the model, it is necessary to prepare a labeled training data set, and the labeled content includes the health status or disease type of acupoints. Select the cross-entropy loss function to optimize the classification task of the model. The formula of the cross-entropy loss function is as follows:

[0031]

[0032] where N represents the total number of categories, y i is the true label, represented by one-hot encoding, the position corresponding to the target category is 1, and the rest are 0, is the probability value predicted by the model, calculated from the logits (unnormalized scores) output by the model through the softmax function. The model is trained through backpropagation and an optimization algorithm (Adam) until the model converges.

[0033] Preferably, for model inference, the preprocessed acupoint coordinates and temperature dictionary are input into the trained Transformer model. The model extracts features through the multi-head self-attention mechanism, fuses global information, and outputs an auxiliary diagnosis result according to the task requirements.

[0034] To ensure the accuracy and generalization ability of the model, a validation set is used to evaluate the model. According to the validation results, the hyperparameters of the model, such as the learning rate, number of layers, number of heads, etc., are adjusted. In addition, methods such as ensemble learning and model pruning can be used to further optimize the model performance to improve the reliability of the diagnosis results.

[0035] In step S6, before the patient is diagnosed, the cloud diagnosis system requires the patient to fill in personal basic information, including name, age, gender, medical history, etc.; the patient can directly select a designated doctor for diagnosis. If the patient does not clearly specify a doctor, the system will intelligently match a professional doctor related to the patient's health status label (such as condition, symptoms, etc.); this matching mechanism is based on the patient's label information, such as the abnormal area shown in the temperature map, meridian symptoms, etc., and combines the doctor's expertise field (for example, acupuncture, massage, internal medicine, etc.) to automatically recommend the most suitable doctor to receive the patient; through this function, the patient can obtain a quick and accurate doctor match, ensuring the efficiency and professionalism of the diagnosis and treatment process;

[0036] Advantages of the present invention:

[0037] Combined with artificial intelligence analysis and based on the temperature changes of each acupoint in the temperature map, the cloud diagnosis system can analyze the temperature distribution of different meridians, help doctors identify possible abnormal areas, and further assist in judging possible health problems of the patient;

[0038] Through the cloud diagnosis system, the generated meridian temperature map, artificial intelligence-assisted diagnosis result and relevant patient health data are automatically synchronized and uploaded. Doctors can access the temperature map and relevant clinical data in real time to make a timely diagnosis of the patient's health status; at the same time, doctors can archive the patient's temperature map data and treatment records for a long time to form a complete patient health file; doctors can view and update the file at any time for follow-up management and long-term health monitoring. Brief Description of the Drawings

[0039] Figure 1 It is a flowchart of an auxiliary diagnosis and processing method based on artificial intelligence and human body image acupoints;

[0040] Figure 2 It is a system block diagram of an auxiliary diagnosis and processing method based on artificial intelligence and human body image acupoints. Detailed Embodiments

[0041] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0042] Embodiment: Currently, in the process of traditional Chinese medicine diagnosis and health management, the analysis of the temperature distribution of human acupoints has important clinical significance. Traditional temperature detection methods usually lack refined processing of the temperature changes in the acupoint area, resulting in difficulty in accurately extracting and presenting the temperature characteristics of meridians and acupoints, thereby affecting the accurate judgment of the doctor on the patient's health status. In traditional Chinese medicine diagnosis, it is difficult to comprehensively reflect the human health status solely relying on temperature data or image data. Although traditional machine learning algorithms or artificial intelligence algorithms can provide certain assistance, in complex environments, the accuracy of existing methods still needs to be improved. To overcome these deficiencies, the present invention proposes an auxiliary diagnosis system based on artificial intelligence and human image acupoints. By combining multi-dimensional temperature matrix data with traditional Chinese medicine meridian theory, it intelligently analyzes the temperature distribution in the acupoint area, generates a visual meridian temperature map, and the artificial intelligence analyzes the temperature information of acupoint coordinates to generate an auxiliary diagnosis result. The system also introduces a cloud intelligent matching mechanism. By analyzing the health data tags of patients and the professional fields of doctors, it realizes intelligent diagnosis and treatment matching, thereby improving the diagnosis efficiency and accuracy.

[0043] As Figure 1 - Figure 2 shown, an auxiliary diagnosis processing method based on artificial intelligence and human image acupoints includes the following steps:

[0044] S1. Receive the temperature matrix generated by the temperature sensor array and the dictionary array containing acupoint coordinates and name information;

[0045] The temperature matrix is a multi-dimensional matrix generated by the temperature sensor array. The temperature matrix records the temperature values of each point in the area, and each matrix element corresponds to the temperature value of a spatial coordinate point. The dictionary stores the coordinate and name information of multiple acupoints.

[0046] S2. For each acupoint, set a radius range with it as the center, obtain the temperature values of all points within the range and calculate the average temperature to generate corresponding regional temperature data;

[0047] 1. After the system receives the temperature matrix data and the acupoint dictionary, parse the coordinates (Xi, Yi) of each acupoint in the dictionary, and at the same time set an effective radius R; traverse all elements of the temperature matrix, and calculate the Euclidean distance between each point and the acupoint by:

[0048]

[0049] Where L is the distance between the matrix element and the acupoint, Xi and Yi are the horizontal and vertical coordinate values of different acupoints, Xn and Yn are the horizontal and vertical coordinate values of a certain matrix element in the temperature matrix, i is the number of acupoints, and n is the number of matrix elements;

[0050] 2. Determine whether each element is within the radius R. If the condition (L ≤ R) is satisfied, then determine that this matrix element is a valid element for calculating the temperature value of a certain acupoint, regard the temperature value of this point as a valid value and include it in the calculation scope; otherwise, exclude this point, that is, if L > R, determine that this matrix element is an invalid element for calculating the temperature value of a certain acupoint.

[0051] 3. Calculate the average value of all valid temperature values within each acupoint. The calculation formula is:

[0052]

[0053] Where Ti is the regional temperature average value of a certain acupoint, Tx is the temperature value of each valid matrix element, and N is the total number of valid elements.

[0054] S3. Precisely associate the acupoint names with the corresponding temperature characteristics to form a structured acupoint - coordinate - temperature relationship dictionary model;

[0055] For each acupoint name and its corresponding coordinates (Xi, Yi) in the new dictionary, generate a new mapping relationship through the temperature average value calculated in step S2, where this acupoint name is directly associated with its regional temperature value;

[0056] The newly generated dictionary structure is: acupoint name: coordinates Xi, Yi, temperature Ti, and the temperature value Ti can be stored with different precisions according to the actual needs of different parts. Support decimal precision control (for example, retaining two decimal places) to meet the refined diagnosis requirements.

[0057] S4. Based on traditional Chinese medicine meridian theory, arrange the acupoints in different meridian systems in a specific order, and generate a precise meridian temperature visualization map according to their temperature characteristics;

[0058] Classify all known meridians, determine the specific acupoints included in each meridian, sort them according to the order of acupoints and their temperature values in each meridian to ensure that the temperature data matches the physiological characteristics of the meridians;

[0059] Match the temperature values of each acupoint in the meridian with its position relationship on the human body to ensure that the temperature value of each acupoint can accurately reflect the physiological state of its location, and generate corresponding temperature layers in order according to the temperature values;

[0060] The temperature map of each meridian is visualized using a suitable color mapping method, such as a heat map and a gradient color bar, to ensure that the regions corresponding to different temperature values are clearly distinguishable; this helps doctors intuitively understand the temperature status of different acupoints and their possible health problems;

[0061] According to the diagnostic requirements and actual situation, adjust the display method of the temperature map of each meridian, including smoothing, refinement, and removal of outliers of temperature data. Further improve the accuracy and practicality of the chart.

[0062] S5. Use the Transformer network to assist in the diagnosis of the acupoint coordinate temperature dictionary and generate the diagnosis results of the artificial intelligence model; it can include disease types, health status, and recommended treatment plans;

[0063] First, organize and format the input acupoint coordinates and temperature data, convert the data into a dictionary format, and normalize the coordinate and temperature data, respectively scaling them to the interval [0, 1] to improve the training efficiency of the model. Finally, embed the normalized data into a high-dimensional space to prepare for the input of the Transformer model;

[0064] Dictionary formatting, for example: {"acupoint 1": (coordinate x1, y1, temperature temp1), "acupoint 2": (coordinate x2, y2, temperature temp2),...}

[0065] Design a Transformer model. Through the multi-head self-attention mechanism, the Transformer network can efficiently process sequence data, capture global dependencies, and does not rely on the sequentiality of the sequence; its parallel processing ability significantly improves the computational efficiency. At the same time, the design of positional encoding retains the sequential information of the sequence, giving it significant advantages in long sequence modeling and complex relationship extraction. Importantly, the introduction of the self-attention mechanism enables the model to dynamically focus on important information at different positions in the sequence, enhancing the ability to capture global dependencies, significantly improving the expressive ability and processing efficiency of the model, especially suitable for long sequence modeling, and applicable to various tasks such as natural language processing and time series analysis.

[0066] The encoder part uses the multi-head self-attention mechanism (Multi-Head Self-Attention) to extract the global dependencies between acupoint coordinates and temperatures. At the same time, positional encoding (Positional Encoding) is introduced to retain the spatial information of the acupoints. In each encoder layer, a feed-forward network and layer normalization operations are added to enhance the expressive ability and stability of the model.

[0067] To train the model, a labeled training dataset needs to be prepared, and the annotation content includes the health status or disease type of acupoints. The cross-entropy loss function is selected to optimize the classification task of the model. The formula of the cross-entropy loss function is as follows:

[0068]

[0069] where N represents the total number of categories, y i is the true label, represented by one-hot encoding, with the corresponding position of the target category being 1 and the rest being 0. is the probability value predicted by the model, calculated from the unnormalized scores of the logits output by the model through the softmax function. The model is trained through backpropagation and optimization algorithms (such as Adam) until the model converges.

[0070] Perform model inference. Input the preprocessed acupoint coordinates and temperature dictionary into the trained Transformer model. The model extracts features through the multi-head self-attention mechanism and fuses global information. According to the task requirements, the model outputs auxiliary diagnosis results, including disease type, health status, and recommended treatment plans.

[0071] To ensure the accuracy and generalization ability of the model, a validation set is used to evaluate the model. According to the validation results, adjust the hyperparameters of the model, such as learning rate, number of layers, number of heads, etc. In addition, the model performance can be further optimized through methods such as ensemble learning and model pruning to improve the reliability of the diagnosis results.

[0072] S6. Connect the generated meridian temperature map with the doctor's diagnosis requirements through the cloud diagnosis system to assist the doctor in conducting health analysis and treatment suggestions based on the temperature chart.

[0073] 1. Before diagnosis, the system requires the patient to fill in personal information, including name, age, gender, medical history, etc. If the patient directly selects a designated doctor, the system automatically matches; if not, the system makes an intelligent match between the patient's health status labels (such as abnormal temperature areas, meridian symptoms) and the doctor's expertise fields (such as acupuncture, massage, internal medicine, etc.) to recommend the most suitable doctor for consultation.

[0074] 2. Based on the patient's meridian temperature map, the cloud diagnosis system analyzes possible health problems. For example, abnormal high-temperature areas may reflect inflammation, and low-temperature areas may indicate blood circulation problems. The system combines meridian sorting and temperature distribution data to generate diagnosis suggestions for the doctor.

[0075] 3. The cloud diagnosis system automatically synchronizes and uploads the generated meridian temperature map with the relevant patient health data. Doctors can access the temperature map and relevant clinical data in real time to make timely diagnoses of the patient's health condition. At the same time, doctors can archive the patient's temperature map data and treatment records in the long term to form a complete patient health file. Doctors can view and update the file at any time for follow-up management and long-term health monitoring.

[0076] Through the use of the temperature matrix generated by the temperature sensor array, the present invention precisely associates the acupoint coordinates in the human body image with the temperature information, constructs regional temperature data and a structured acupoint-temperature relationship model. Based on traditional Chinese medicine meridian theory, it classifies and sorts the acupoint temperatures in the meridians, generates a visualized meridian temperature map, and intuitively displays the human body temperature state in combination with visualization methods such as heat maps. In addition, a diagnosis result prediction is carried out through the Transformer network. Through the cloud diagnosis system, the present invention realizes the precise docking and real-time analysis of the patient temperature map, diagnosis result prediction and health file, effectively assisting doctors in making health diagnoses and treatment suggestions, making up for the technical gap of the lack of complete acupoint temperature curves in the traditional meridian system and the problem that the accuracy still needs to be improved, and improving the accuracy and efficiency of diagnosis and treatment.

[0077] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, for those skilled in the art, they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An auxiliary diagnosis and processing method based on artificial intelligence and human body image acupoints, characterized in that: The following steps are involved: S1, receiving a temperature matrix generated by a temperature sensor array and a dictionary array containing acupoint coordinates and name information; S2. For each acupuncture point, a radius range is set with the acupuncture point as the center, the temperature values ​​of all points within the range are obtained and the average temperature is calculated to generate the corresponding regional temperature data; S3, accurately associating the acupoint names with the corresponding temperature characteristics to form a structured acupoint-coordinate-temperature relationship dictionary model; S4. Based on the meridian theory of traditional Chinese medicine, the acupoints in different meridian systems are arranged in a specific order, and accurate meridian temperature visualization maps are generated according to their temperature characteristics; S5, based on the Transformer network, assist in the diagnosis of the acupoint coordinate temperature dictionary and generate the artificial intelligence model diagnosis results; S6. The generated meridian temperature map is connected with the doctor's diagnosis needs through the cloud diagnosis system to assist the doctor in conducting health analysis and treatment recommendations based on the temperature map.

2. The auxiliary diagnosis and processing method based on artificial intelligence and human body image acupoints according to claim 1 is characterized in that: The temperature matrix in step S1 is a multidimensional matrix generated by a temperature sensor array. The temperature matrix records the temperature values ​​of each point in the area. Each matrix element corresponds to the temperature value of a spatial coordinate point. The dictionary stores the coordinates and name information of multiple acupoints.

3. The auxiliary diagnosis and processing method based on artificial intelligence and human body image acupoints according to claim 1 is characterized in that: The calculation method of step S2 is as follows: Where L is the distance between the matrix element and the acupuncture point, Xi, Yi are the horizontal and vertical coordinate values ​​of different acupuncture points, Xn, Yn are the horizontal and vertical coordinate values ​​of a matrix element in the temperature matrix, i is the number of acupuncture points, and n is the number of matrix elements; If L￡R, the matrix element is determined to be a valid element for calculating the temperature value of a certain acupoint point; if L>R, the matrix element is determined to be an invalid element for calculating the temperature value of a certain acupoint point; Where Ti is the regional temperature average of a certain acupuncture point, Tx is the temperature value of each valid matrix element, and N is the total number of valid elements.

4. The auxiliary diagnosis and processing method based on artificial intelligence and human body image acupoints according to claim 1 is characterized in that: The step S3 comprises: For each acupoint name and its corresponding coordinates (Xi, Yi) in the new dictionary, a new mapping relationship is generated by the average temperature value calculated in step S2, in which the acupoint name is directly associated with its regional temperature value; The newly generated dictionary structure is: acupoint name: coordinates Xi, Yi, temperature Ti, and the temperature value Ti can be stored with different precisions according to the actual needs of different parts.

5. The auxiliary diagnosis and processing method based on artificial intelligence and human body image acupoints according to claim 1 is characterized in that: The step S4 comprises: Classify all known meridians, determine the specific acupoints contained in each meridian, and sort them according to the order of acupoints in each meridian and their temperature values ​​to ensure that the temperature data matches the physiological characteristics of the meridian; Match the temperature values ​​of each acupuncture point in the meridian with its position on the human body to ensure that the temperature value of each acupuncture point can accurately reflect the physiological state of its location, and generate corresponding temperature layers in sequence; The temperature map of each meridian adopts appropriate color mapping, using heat map and gradient color bar visualization methods to ensure that the areas corresponding to different temperature values ​​can be clearly identified; According to diagnostic needs and actual conditions, the display mode of each meridian temperature map is adjusted, including the refined processing of temperature data and the removal of abnormal values.

6. The auxiliary diagnosis and processing method based on artificial intelligence and human body image acupoints according to claim 1 is characterized in that: The step S5 firstly organizes and formats the input acupoint coordinates and temperature data, organizes the data into a dictionary format, normalizes the coordinates and temperature data, and scales them to the [0, 1] interval respectively. Finally, the normalized data is embedded into a high-dimensional space to prepare for the input of the Transformer model. The encoder part of the Transformer is designed, and a multi-head self-attention mechanism is used to extract the global dependency between acupoint coordinates and temperature. At the same time, position encoding is introduced to retain the spatial information of the acupoints. In each encoder layer, a feedforward network and layer normalization operations are added.

7. The auxiliary diagnosis and processing method based on artificial intelligence and human body image acupoints according to claim 6 is characterized in that: In order to train the model, it is necessary to prepare a labeled training data set, which includes the health status or disease type of the acupoints. The cross entropy loss function is selected to optimize the classification task of the model. The formula is as follows: Where N represents the total number of categories, y i is the true label, represented by one-hot encoding, the corresponding position of the target category is 1, and the rest are 0. It is the probability value predicted by the model, which is calculated from the unnormalized score of the logits output by the model through the softmax function. The model is trained through back propagation and optimization algorithms until the model converges.

8. The auxiliary diagnosis and processing method based on artificial intelligence and human body image acupoints according to claim 7 is characterized in that: Perform model inference and input the preprocessed acupoint coordinates and temperature dictionary into the trained Transformer model. The model extracts features through a multi-head self-attention mechanism and fuses global information. According to task requirements, the model outputs auxiliary diagnosis results.

Images