A multi-model collaborative decision method and system for digestive tract image analysis
By constructing a heterogeneous dual-branch feature extraction structure combining convolutional neural networks and visual Transformers, and combining it with a learnable dynamic weighted fusion mechanism, the shortcomings of extracting local fine-grained and global structural information in gastrointestinal image analysis are solved, improving the accuracy and stability of lesion identification. It is suitable for automatic analysis and assisted diagnosis of gastroscopy, colonoscopy and capsule endoscopy images.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- XUZHOU FIRST PEOPLES HOSPITAL
- Filing Date
- 2026-05-13
- Publication Date
- 2026-07-17
AI Technical Summary
Existing methods for analyzing digestive tract images are insufficient in extracting local fine-grained lesion features and global structural semantic information. They are unable to take into account both minor lesions and overall structural features, and their generalization ability is insufficient due to differences across devices and imaging conditions.
A heterogeneous dual-branch feature extraction structure based on convolutional neural networks and visual Transformers is constructed, and a learnable dynamic weighted fusion mechanism is introduced to achieve collaborative modeling of local fine-grained features and global semantic information. The recognition accuracy and stability are improved by adaptively adjusting the decision weights.
It improves the accuracy of lesion identification and cross-scenario stability in gastrointestinal image analysis, enhances the model's ability to express complex lesions and its robustness, and has good scalability and practical application value.
Smart Images
Figure CN122415573A_ABST