An over-segmentation multi-instance learning glioma gene heterogeneity visualization method
CN122290912APending Publication Date: 2026-06-26FUDAN UNIVERSITY
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- FUDAN UNIVERSITY
- Filing Date
- 2024-12-26
- Publication Date
- 2026-06-26
AI Technical Summary
Technical Problem
Current technologies make it difficult to perform quantitative assessments of gene heterogeneity within gliomas non-invasively, leading to inconsistent surgical planning and treatment outcomes.
Method used
By employing a superpixel multi-instance learning method, tumor image patches are extracted and clustered into superpixel regions. Combined with radiomics features, a dual attention branch model is constructed to visualize and quantify tumor IDH1 gene heterogeneity.
Benefits of technology
It enables non-invasive quantitative assessment of IDH1 gene heterogeneity in gliomas, providing guidance for surgical planning and targeted therapy, and improving the uniformity of treatment outcomes.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN122290912A_ABST
Abstract
This invention relates to a superpixel multi-instance learning method for visualizing glioma gene heterogeneity. The method includes the following steps: Step 1, extracting tumor image patches from an image to obtain a superpixel map output by a pre-trained ResNet50 model; Step 2, extracting global high-throughput features of the tumor from the image to obtain radiomics features; Step 3, training a superpixel multi-instance network to obtain a superpixel multi-instance model; Step 4, acquiring actual case images, inputting them into the superpixel multi-instance model, and obtaining the final IDH1 heterogeneity distribution visualization map. Compared with existing technologies, this invention has advantages such as achieving quantification of IDH1 heterogeneity across the entire tumor region.
Need to check novelty before this filing date? Find Prior Art