Pathological image classification method and system based on multi-scale domain adversarial network

Abstract

One technical scheme of the invention provides a pathological image classification method based on a multi-scale domain adversarial network. Another technical scheme of the invention is to provide a pathological image classification system based on a multi-scale domain adversarial network, which is characterized by comprising a preprocessing module; a single-scale feature extraction module; an overall feature extraction module; a multi-scale attention module; and a model evaluation module. According to the method, on one hand, WSI multi-scale feature information is combined, on the other hand, the influence of different dyeing effects on a prediction result is inhibited by using the domain adversarial network, and the volatility of the pathological image caused by dyeing is reduced, so that a system for assisting the pathological doctor to classify the pathological image in a manner of simulating the actual operation process of the pathological doctor is provided.

Description

technical field [0001] The invention relates to a pathological image classification method and system based on a multi-scale domain confrontation network, belonging to the field of medical image processing. Background technique [0002] Computer-aided diagnosis technology based on artificial intelligence has been widely used in the medical field, especially in the diagnosis of histopathological slides (Whole Slide Image, hereinafter referred to as "WSI"). Using the automatic recognition technology of deep learning to intelligently analyze WSI can assist pathologists to complete pathological analysis work efficiently and accurately. At present, cancer classification tasks for digital pathology images are mainly implemented based on weakly supervised deep learning. For example, Chinese patent CN202010690425.5, Chinese patent CN202010287157.2 and Chinese patent CN201910120656.X all use the label of the entire WSI as the package label. The tiles segmented at the maximum resolut...

AI Technical Summary

Problems solved by technology

At present, cancer classification tasks for digital pathology images are mainly implemented based on weakly supervised deep learning. For example, Chinese patent CN202010690425.5, Chinese patent CN202010287157.2 and Chinese patent CN201910120656.X all use the label of the entire WSI as the package label. The tiles segmented at the maximum resolution are used as instances in the package to construct a multi-instance learning framework and complete the classification task of pathological images in a weakly supervised learning manner, but they do not consider many in the whole prediction process. The impact of scale feature information and coloring bias on classification results

Images