An artificial intelligence-based auxiliary diagnosis method for digestive tract diseases

By constructing a knowledge graph of digestive tract diseases and a deep learning model, and optimizing feature extraction and attention mechanisms, the problems of inaccurate feature selection and lack of medical knowledge constraints in endoscopic image analysis were solved, thus achieving efficient and accurate diagnosis of digestive tract diseases.

CN122117334APending Publication Date: 2026-05-29BEIJING FRIENDSHIP HOSPITAL CAPITAL MEDICAL UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
BEIJING FRIENDSHIP HOSPITAL CAPITAL MEDICAL UNIV
Filing Date
2026-03-03
Publication Date
2026-05-29

AI Technical Summary

Technical Problem

Existing endoscopic image analysis methods are inaccurate in feature selection and lack medical knowledge constraints in the diagnosis of gastrointestinal diseases, resulting in high misdiagnosis rates and low diagnostic efficiency.

Method used

By constructing a knowledge graph of digestive tract diseases to generate a constraint matrix, and combining multi-feature extraction and deep learning models, the cross-modal attention mechanism is optimized to extract features such as mucosal texture complexity and vascular morphology distribution entropy for disease prediction.

Benefits of technology

It improves the accuracy of feature selection, enhances the interpretability of the model, reduces the misdiagnosis rate, and improves diagnostic efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122117334A_ABST
    Figure CN122117334A_ABST
Patent Text Reader

Abstract

The application belongs to the technical field of data processing and artificial intelligence, and mainly relates to an auxiliary diagnosis method for digestive tract diseases based on artificial intelligence, comprising the following steps: first, a knowledge graph of digestive tract diseases is constructed, and a constraint matrix is constructed based on entity information and relationship information in the graph; then, image data is acquired by using NBI endoscope images, the data is standardized, and mucosal texture complexity (MTC) features, vessel morphology distribution entropy (VMDE) features, ulcer area fractal dimension (UFD) features, mucosal layer thickness change rate (MTCR) features and multi-spectral reflectance distribution (MSRD) features are extracted from the processed image data; then, the extracted features are constructed into a feature vector, and a constrained feature vector is obtained through constraint operation with the constraint matrix; finally, the constrained feature vector is input into a deep learning neural network model, and a final diagnosis result is output; this method not only improves the accuracy of feature selection, but also reduces the misdiagnosis rate.
Need to check novelty before this filing date? Find Prior Art