Metalearning-based few-sample medical microscopic hyperspectral image segmentation method

By combining a hybrid convolution module, a multi-scale feature extractor, and a learnable edge detection module, the problems of single feature extraction and insufficient multi-scale adaptation in medical hyperspectral image segmentation are solved, achieving accurate segmentation and robustness improvement of lesion tissues, and adapting to rapid generalization in small sample scenarios.

CN122176309APending Publication Date: 2026-06-09HOHAI UNIV
0 Cites 1 Cited by

Patent Information

Application Number
CN202610329037.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-03-18
Publication Date
2026-06-09

AI Technical Summary

Technical Problem

Existing medical hyperspectral image segmentation methods suffer from limitations such as single feature extraction, insufficient multi-scale adaptation, difficulty in capturing weak edges, and lack of adaptive capability for small samples, making it difficult to achieve accurate segmentation of lesions and normal tissues.

Method used

By employing the synergistic effect of a hybrid convolution module, a multi-scale feature extractor, a learnable edge detection module, and a customized meta-learning framework, the hybrid convolution module adaptively selects the convolution type, and by combining multi-scale feature fusion and the meta-learning framework, it achieves adaptation to distorted features and accurate capture of weak edges, enabling rapid generalization in small sample scenarios.

Benefits of technology

It achieves precise segmentation of diseased tissue, improves segmentation accuracy and robustness, adapts to different lesion morphologies and equipment differences, and provides reliable clinical diagnostic support.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention discloses a medical hyperspectral image segmentation method based on hybrid convolution fusion meta learning, and relates to the technical field of medical image processing and deep learning. In order to solve the technical defect that an existing medical hyperspectral segmentation algorithm is difficult to adapt to target deformation, multi-scale difference, weak edge features and small sample scenes, the method realizes accurate segmentation of a lesion area through an integrated process of data preprocessing, mixed feature extraction, multi-scale fusion, learnable edge optimization and meta-learning self-adaption. The method comprises the following steps: firstly, carrying out preprocessing such as screening, labeling and enhancing on a medical hyperspectral image; multi-type features are adaptively extracted through a hybrid convolution module, and key information is enhanced through a channel space attention module; fusing deep and shallow layer features by using a multi-scale feature extractor to adapt to lesions of different scales; accurately capturing weak edges through a learnable edge detection module; and finally, in combination with a customized meta-learning framework, new data distribution can be quickly adapted only by a small number of samples. Experimental results show that the method has excellent performance in gastric cancer and bile duct cancer hyperspectral image segmentation tasks and stable cross-dataset generalization performance, the reliability and robustness of the method are proved, and a quantitative basis is provided for early diagnosis of clinical diseases.
Need to check novelty before this filing date? Find Prior Art