基于多模态影像动态特征融合的肿瘤分割方法
By employing a multimodal image dynamic feature fusion method, utilizing a tumor boundary manifold model and nonlinear feature weight adjustment, the problems of blurred tumor boundaries and complex morphology in single-modal image segmentation are solved, thereby improving the accuracy and reliability of tumor segmentation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- YANGZHOU UNIV
- Filing Date
- 2025-05-29
- Publication Date
- 2026-07-17
AI Technical Summary
Single-modal medical image segmentation methods are difficult to achieve ideal segmentation results due to the blurred boundaries and complex morphology of tumors, resulting in reduced accuracy and reliability of image segmentation.
A multimodal image dynamic feature fusion method is adopted. By acquiring multimodal medical image data, a heat distribution probability map is generated using a tumor boundary manifold model. The convolution kernel parameters are dynamically adjusted, nonlinear feature weights are adjusted, and the multimodal tumor boundary features are fused to generate tumor segmentation results.
It improves the accuracy of tumor boundary recognition and the accuracy and reliability of image segmentation, especially maintaining good segmentation results even under low-quality images, and enhances the ability to suppress noise.
Smart Images
Figure CN120580245B_ABST