Medical image full-automatic segmentation method based on multi-scale attention mechanism network
By using adaptive sliding window segmentation and Gaussian space weighted fusion through a multi-scale attention mechanism network, the problem of balancing segmentation accuracy and efficiency in large-size medical image segmentation is solved, achieving high-precision, low-cost fully automatic segmentation, which is suitable for clinical applications of multimodal medical images.
CN122415997APending Publication Date: 2026-07-17SICHUAN YUNTONG ZHILIAN TECHNOLOGY CO LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SICHUAN YUNTONG ZHILIAN TECHNOLOGY CO LTD
- Filing Date
- 2026-05-25
- Publication Date
- 2026-07-17
Smart Images
Figure CN122415997A_ABST
Abstract
本发明公开了一种基于多尺度注意力机制网络的医学影像全自动分割方法，其属于医学影像分割技术领域，其解决了现有技术在应对大尺寸医学影像分割时存在难以同时兼顾分割精度与计算效率的问题。本发明通过粗分割定位和自适应滑窗精细分割的两级架构，先以轻量化网络快速确定感兴趣区域，再根据目标分布自适应调整滑窗尺寸与重叠率，密集区域用小窗高重叠保证精度，稀疏区域用大窗低重叠减少冗余，并利用高斯加权融合消除拼接伪影，从而在保留微小病灶细节的同时有效控制计算开销，实现大尺寸医学影像的高精度高效全自动分割。
Need to check novelty before this filing date? Find Prior Art