Tongue Diagnosis Analysis System Based on Image Semantic Coding and Symptom Correlation Validation

Abstract

A tongue diagnosis analysis system based on image semantic coding and symptom association verification is proposed. The system uses an image acquisition module to capture images of the user's front and back of the tongue. An image preprocessing unit performs quality inspection and segmentation to obtain the tongue body region. A symptom association derivation unit converts relevant features of the tongue body region into standardized semantic labels, forming a tongue image semantic code. The system then queries a pathological symptom association database based on each label to generate a list of symptoms with association weights. Through an interactive interface, the user can confirm each symptom to form a confirmed symptom set. This set, along with the tongue image semantic code, is input into a constitution assessment model. The system outputs matching scores for various TCM constitution types and an interpretable tongue diagnosis analysis report, which can be used for long-term tracking and model optimization.

Claims

1. A tongue diagnosis analysis system based on image semantic coding and symptom association verification, comprising: An image capturing module is equipped with a lens for capturing images of the front and back of a user's tongue and stores standardized shooting conditions for correcting shooting parameters; a processor is electrically connected to the image capturing module; a memory is electrically connected to the processor for storing computer program instructions, the processor being configured to control the following unit to execute the computer program instructions: an image preprocessing unit is electrically connected to the processor and the image capturing module for receiving the original tongue image transmitted by the image capturing module, performing image quality detection, correcting shooting parameters and removing background, so as to segment a region of interest on the tongue; A semantic encoding unit, electrically connected to the processor and the image preprocessing unit, is used to execute a feature extraction algorithm to identify local visual features of tongue color, tongue coating color, tongue shape, spots, texture, and sublingual veins in the region of interest of the tongue output by the image preprocessing unit. Based on a predefined set of tongue image semantic labels, these local visual features are mapped to corresponding standardized semantic labels to generate a tongue image semantic code, which is then stored as traceable intermediate layer data. A symptom association derivation unit, electrically connected to the processor and the semantic encoding unit, is used to query a pathological symptom association database stored in the memory, obtain statistically correlated predicted symptoms and corresponding association weights based on the tongue image semantic code, and generate a symptom association list. An interactive interface unit, electrically connected to the processor and the symptom association derivation unit, is used to generate a status confirmation table on a display based on the symptom association list and the association weights, and receive user confirmation input for each predicted symptom to generate a confirmed symptom set. A constitution determination and report generation unit is electrically connected to the processor, the semantic encoding unit, and the interactive interface unit, respectively. It is used to receive the tongue image semantic encoding and the confirmation symptom set output by the semantic encoding unit and the interactive interface unit, convert the tongue image semantic encoding and the confirmation symptom set into input features of the constitution determination model for collaborative authentication, calculate the constitution matching score of at least one TCM constitution type, and output a tongue diagnosis analysis report.

2. The tongue diagnosis analysis system based on image semantic coding and symptom association verification as described in claim 1, wherein, During the shooting process, the image preprocessing unit displays a composition guide frame on a preview screen and detects in real time whether the tongue falls into the composition guide frame and whether the ambient light intensity is higher than a predetermined threshold value. Shooting is only allowed when the composition and lighting conditions are met.

3. The tongue diagnosis analysis system based on image semantic coding and symptom association verification as described in claim 1, wherein, The pathological symptom association database stores the association weights between each standardized semantic label and multiple symptoms. The symptom association derivation unit has an update interface, which is displayed on a monitor through the interactive interface unit to adjust the association weights based on new samples or expert annotations.

4. The tongue diagnosis analysis system based on image semantic coding and symptom association verification as described in claim 1, wherein, The interactive interface unit is configured to set the display order or display priority according to the correlation weight of each predicted symptom, and can automatically collapse predicted symptoms with low correlation weight based on the user's previous answers.

5. The tongue diagnosis analysis system based on image semantic coding and symptom association verification as described in claim 1, wherein, The constitution assessment model is a multi-label classifier that outputs matching scores for multiple constitution types. The constitution assessment and report generation unit determines to display one or more constitution results based on a predetermined threshold or sorting rule.

6. The tongue diagnosis analysis system based on image semantic coding and symptom association verification as described in claim 1, wherein, The semantic encoding unit generates a frontal semantic code and a sublingual semantic code for the frontal image of the tongue and the back image of the tongue, respectively. The constitution determination and report generation unit gives a higher weight to the sublingual semantic code when calculating the blood stasis constitution matching score.

7. The tongue diagnosis analysis system based on image semantic coding and symptom association verification as described in claim 1, wherein, The tongue diagnosis analysis report includes a description of the patient's constitution, key standardized semantic labels that determine the constitution and their corresponding confirmed symptoms, as well as health education recommendations.

8. The tongue diagnosis analysis system based on image semantic coding and symptom association verification as described in claim 1, wherein, The memory further stores the tongue image semantic encoding, confirmed symptom set and constitution judgment results of previous users. The symptom association deduction unit and the constitution judgment and report generation unit can automatically retrain or recalibrate the parameters of the constitution judgment model based on the accumulated historical data.

Images