Pathological slice image segmentation method, device, computer equipment and storage medium

Abstract

The present application relates to a pathological slice image segmentation method, device, computer equipment and storage medium. The method includes: acquiring image blocks of pathological slice images; inputting the image blocks into a first segmentation network to obtain a fine-grained prediction probability map; down-sampling and up-sampling the fine-grained prediction probability map through a first preset magnification The operation obtains a medium-fine-grained prediction probability map; the medium-fine-grained prediction probability map is subjected to second preset magnification downsampling and up-sampling operations to obtain a coarse-grained prediction probability map; the image block, fine-grained prediction probability map, medium The fine-grained prediction probability map and the coarse-grained prediction probability map are concatenated and input to the second segmentation network to obtain a segmentation result map with the same size as the image block. The method can improve the accuracy of the boundary of tissue necrosis area.

Description

technical field [0001] The present application relates to the technical field of image processing, in particular to a pathological slice image segmentation method, device, computer equipment and storage medium. Background technique [0002] In clinical diagnosis, it is necessary to segment the necrotic area and the normal area according to the slices of the pathological site in order to analyze the morphological characteristics of the necrotic area. However, the existing methods can only be segmented manually or through rough image processing. The accuracy of this segmentation is often not high, which affects subsequent diagnosis. Pathological slices are tissue slices of the body, such as liver slices, pancreas slices, and lung slices. For example, pathologists need to manually segment necrotic areas and normal areas in hematoxylin-eosin-stained digital pathological sections of the liver, and then perform quantitative analysis on the morphological characteristics of each nec...

AI Technical Summary

Problems solved by technology

[0002] In clinical diagnosis, it is necessary to segment the necrotic area and the normal area according to the slices of the pathological site in order to analyze the morphological characteristics of the necrotic area. However, the existing methods can only be segmented manually or through rough image processing. This kind of segmentation is often not very accurate, which affects subsequent diagnosis. Pathological slices are tissue slices of the body, such as liver slices, pancreas slices, lung slices, etc.

For example, pathologists need to manually segment necrotic areas and normal areas in hematoxylin-eosin-stained digital pathological sections of the liver, and then perform quantitative analysis on the morphological characteristics of each necrotic area, and finally perform quantitative analysis based on the quantitative analysis results. The patients were graded and diagnosed; the existing liver necrosis area segmentation method can roughly segment the necrosis area in the slice; however, due to the small difference between the liver necrosis area and the normal area, and the boundary is not easy to observe (especially the mild necrosis area), Therefore, these methods are often not accurate enough for the positioning of the boundary.

[0003] The existing pathological slice image segmentation is prone to wrong borders, and wrong border positioning can easily cause two adjacent necrotic areas to be segmented into one, which will affect the final morphological information statistics and cancer grading diagnosis

Images