Pathological image analysis method based on cross-scale spatial constraint fusion and local perception

By combining cross-scale spatial constraint fusion with local perception, and utilizing techniques such as bidirectional cross attention and Hadamard product, the problem of insufficient utilization of multi-resolution information in pathological image analysis is solved, the stability and accuracy of feature representation are improved, and the workload and error rate of manual inspection are reduced.

CN120656169AActive Publication Date: 2025-09-16NANJING UNIV OF INFORMATION SCI & TECH
4 Cites 1 Cited by

Patent Information

Application Number
CN202511174076.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-08-21
Publication Date
2025-09-16
Estimated Expiration
2045-08-21

AI Technical Summary

Technical Problem

Existing pathology image analysis methods fail to fully utilize the instance connections between different spatial scales and ignore multi-resolution information, resulting in insufficient feature aggregation and lack of local information fusion, making manual inspection labor-intensive, time-consuming and error-prone.

Method used

A cross-scale spatial constraint fusion and local perception method is adopted to extract features through a self-supervised pathology basic model, and feature fusion is performed using bidirectional cross attention and Hadamard product. Combined with residual connection and local perception enhancement modules, high-order semantic information is captured and local information is introduced to improve feature representation capabilities.

Benefits of technology

It improves the stability, accuracy and efficiency of pathology image processing, effectively integrates features of different scales, and reduces the workload and error rate of manual inspection.

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

Abstract

The invention provides a pathological image analysis method based on cross-scale spatial constraint fusion and local perception. The method comprises the steps of 1, data preprocessing and feature extraction; step 2, executing a two-way fusion strategy based on spatial constraint to obtain fused enhanced features; step 3, capturing high-order semantic information between scales by using a bilinear fusion mode based on a Hadamard product; 4, performing residual connection on the enhanced features and the fused features to serve as input of a 10-time branch and a 20-time branch; 5, remodeling the features after residual connection into a two-dimensional shape, and inputting three convolutional layers which are arranged in a layered manner and expanded layer by layer; and 6, further performing gating adaptive fusion on the global semantic features obtained after the 10-time branches and the 20-time branches are subjected to the step 5. On the basis of easy deployment, feature representation with high expression ability is obtained, information between branches of different scales is fully utilized, and better performance is obtained in downstream tasks.
Need to check novelty before this filing date? Find Prior Art