Deep migration semi-supervised domain adaptive classification method for histopathology images

Abstract

The invention relates to the technical field of histopathological image classification, in particular to a depth migration semi-supervised domain self-adaptive classification method for histopathological images, which adopts a self-adaptive threshold filtering and sampling strategy to be used for patch dicing so as to reduce time consumption. According to the method, deep migration learning can be realized by migrating a patch-based CNN network HisNet; according to the method, high-order features of a source domain and a target domain are extracted through a reconstruction network, and alignment is carried out in a self-adaptive mode through a semi-supervised domain with multiple weighted loss function criteria; a novel manifold regularization loss function is introduced to fully utilize the characteristics of a target domain sample and obtain a better classification result. Experimental results show that the method provided by the invention has high accuracy, efficiency and stability in histopathology image classification, and has great clinical application value.

Description

technical field [0001] The invention relates to the technical field of classification of histopathological images, in particular to a depth migration semi-supervised domain self-adaptive classification method of histopathological images. Background technique [0002] Histopathological image analysis is very important for cancer diagnosis. In general clinical practice, the classification of cancerous tissue is usually performed by experienced histopathologists by analyzing histopathological images. But the process is highly subjective, time-consuming and prone to misdiagnosis. [0003] Computer-aided diagnosis (CAD) procedures increase diagnostic accuracy and reduce reliance on pathologist experience. Compared with traditional machine learning methods, convolutional neural network (CNN) can automatically learn from raw images and find representative and differentiated information without any preprocessing. Recently, CNNs have been widely used in various medical image diagn...

AI Technical Summary

Problems solved by technology

[0004] The present invention provides a method for self-adaptive classification of histopathological images in the depth migration semi-supervised domain, which solves the technical problem of how to obtain good classification results in the case of limited labeled samples

Images