Method for establishing brain glioma grading hybrid network based on multi-view feature fusion
By constructing a hybrid network for glioma grading that integrates multi-view features, and utilizing multi-view MRI data and an innovative feature extraction module, the problems of time-consuming, labor-intensive, and information-loss-prone traditional imaging diagnostic methods are solved, enabling comprehensive understanding and efficient classification of gliomas.
Patent Information
- Application Number
- CN202411758380.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-02
- Publication Date
- 2026-04-10
- Estimated Expiration
- 2044-12-02
AI Technical Summary
Traditional imaging diagnostic methods rely on physician experience, are time-consuming and labor-intensive, and are highly subjective. Single-modal MRI analysis cannot capture the heterogeneity and complexity within tumors, while multimodal MRI fusion methods ignore scanning planes, leading to information loss. Attention mechanisms are not good at extracting local features, affecting the accuracy and reliability of classification results.
A hybrid network for glioma grading based on multi-view feature fusion is constructed. Through data preprocessing, feature extraction and multi-view feature fusion, MRI data from inter-slice, spatial and modal perspectives are used. Global, local and multi-scale feature extraction is performed by combining focused overlapping spatial reduction attention and adaptive dilated convolution. Feature expression is enhanced by a multi-view feature fusion module.
This approach enables a comprehensive understanding of gliomas, balances the global receptive field and inductive bias, enhances feature extraction capabilities, and improves the accuracy and reliability of classification results.