Full-slice glomerular image segmentation method and device, electronic equipment and storage medium

Through adaptive cropping and image segmentation model based on nnU-Net network, the full-sliced ​​glomerulus image is segmented, which solves the problem of low image segmentation accuracy in the prior art and achieves higher accuracy and robustness.

CN119722700APending Publication Date: 2025-03-28TSINGHUA UNIVERSITY
3 Cites 0 Cited by

Patent Information

Application Number
CN202411716850.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-27
Publication Date
2025-03-28

AI Technical Summary

Technical Problem

The existing glomerular image segmentation method cannot fully capture the characteristics of different anatomical structures and lesion areas in renal tissue, resulting in low image segmentation accuracy.

Method used

By acquiring the original fully sliced ​​glomerulus image and its corresponding optical scaling magnification, adaptive cropping is performed to obtain multiple input image blocks, and these image blocks are segmented based on the pre-trained image segmentation model, and finally the segmentation results are overlapped and spliced ​​to obtain the target segmentation result.

Benefits of technology

Accurate segmentation of complex anatomical structures and lesion areas in full-section glomerulus images is achieved, improving the accuracy and robustness of segmentation results.

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

Abstract

The invention provides a full-slice glomerular image segmentation method and apparatus, an electronic device and a storage medium. The method comprises the steps of obtaining an original full-slice glomerular image and a corresponding optical zoom ratio thereof; according to an optical zoom ratio and an edge removal mechanism, carrying out adaptive cutting on the original full-slice glomerular image to obtain a plurality of input image blocks; segmenting the plurality of input image blocks one by one based on a pre-trained image segmentation model to obtain a plurality of image block segmentation results; wherein the image segmentation model is constructed based on an nnU-Net network and a variant thereof; and overlapping and splicing the plurality of image block segmentation results to obtain a target segmentation result of the original full-slice glomerular image. According to the method, through priori knowledge based on the optical zoom ratio, it is ensured that the scale of the cut input image blocks is consistent with that of the training image, meanwhile, by introducing an edge removal mechanism and adopting an overlapping mode for cutting, the edge effect when the glomerulus is located at the cut edge is effectively avoided, and the accuracy of the cut glomerulus is improved. Accurate segmentation of a complex anatomical structure and a lesion area in the full-slice glomerular image is realized, and the accuracy and robustness of an image segmentation result are improved.
Need to check novelty before this filing date? Find Prior Art