Semi-supervised learning cell segmentation method based on a generative adversarial network

Abstract

The invention discloses a semi-supervised learning cell segmentation method based on a generative adversarial network, which comprises the following steps: collecting cell segmentation data, preprocessing and enhancing the data, and dividing the data into a training set and a test set picture. A new adversarial generation network is designed by taking semi-supervised learning as a starting point.Compared with a previous adversarial generation network, the network replaces a generator with a small-parameter full-volume integral cut network and is used for outputting a probability graph to an input picture. For a cell picture without a label, a semi-supervised method is used for training a segmentation network, after initial segmentation prediction of an unmarked image is obtained from thesegmentation network, a segmentation prediction probability graph is transmitted through a discrimination network, and a confidence graph is obtained. The confidence map is used as a supervision signal, a self-learning mechanism is used to train a segmentation network, and the confidence map represents the quality of prediction segmentation. Through the convolutional neural network designed by theinvention, the cell segmentation accuracy is improved.

Description

technical field [0001] The invention relates to the field of biomedical image processing and computer application, in particular to a cell segmentation method based on semi-supervised learning of confrontation generation network. Background technique [0002] Cell segmentation is the most important step in the study of cell movement and cell morphology. Segmenting precise cells from medical images is currently a challenging topic. With the rapid development of Internet technology, traditional research methods have become less applicable. The traditional method of using human eyes under a microscope to stain, classify, and count cells requires a lot of cumbersome human operations, and its reusability relatively low. [0003] In the research of machine vision, segmentation methods based on deep learning can effectively solve some object segmentation problems. But the most important prerequisite is that a large number of manually labeled medical images are required, which req...

AI Technical Summary

Problems solved by technology

[0009] The technical problem to be solved by the present invention is to provide a method to solve the problem of relying on a large amount of manual labeling ...

Images