一种基于多源数据驱动的分层动态融合病灶识别系统
The hierarchical dynamic fusion lesion identification system driven by multi-source data utilizes multimodal feature extraction and fusion technology to solve the problems of lack of pathological priors, non-targeted feature extraction, and weak scene adaptability in existing methods. It achieves a significant improvement in the accuracy and comprehensiveness of lesion identification and is suitable for clinical applications involving multiple organs and various lesion types.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- THE THIRD MEDICAL CENT OF THE CHINESE PEOPLES LIBERATION ARMY GENERAL HOSPITAL
- Filing Date
- 2026-05-11
- Publication Date
- 2026-07-17
AI Technical Summary
Existing multimodal medical image fusion methods do not incorporate prior pathological knowledge, lack targeted feature extraction, lack scene adaptability, are difficult to adapt to small sample scenarios, and do not fully explore the complementarity of features of each modality, resulting in insufficient accuracy and comprehensiveness in lesion identification.
A hierarchical dynamic fusion lesion identification system based on multi-source data is adopted. Through CT feature extraction module, MRI feature extraction module, PET feature extraction module and ultrasound feature extraction module, combined with multi-task learning module and fusion module, the system uses adaptive spatial pyramid, attention-enhanced frequency-space dual-stream network, graph convolutional network and time-aware recurrent network and other technologies to extract the core features of each modality, and performs lesion identification through hierarchical attention and meta-learning.
It improves the accuracy and comprehensiveness of lesion identification, is compatible with multiple organs and lesion types, is suitable for clinical auxiliary diagnosis and lesion screening, solves the problems of missed diagnosis and misdiagnosis of lesions, and is adapted to the actual clinical needs of small samples and multiple scenarios.
Smart Images
Figure CN122156885B_ABST