Gastric lesion segmentation method and system based on generative adversarial network

Abstract

The invention discloses a stomach lesion segmentation method and system based on a generative adversarial network, and the method comprises the steps: inputting a gastroscope lesion picture sample into a segmentation network, and obtaining a segmentation prediction image of a lesion region; wherein the segmentation network comprises an encoder, an expansion convolution module and a decoder based on U-Net; a four-channel tensor formed by splicing a gastroscope lesion image sample and a segmentation prediction image and a four-channel tensor formed by splicing a gastroscope image and a manual annotation image are used as two groups of inputs of a discrimination network, and whether the two groups of inputs are the segmentation prediction image or the manual annotation image is judged as outputs. Alternately performing generative adversarial network training on the segmentation network and the discrimination network in a game mode to a balance state; and inputting a to-be-segmented gastroscope picture into the trained segmentation network to obtain a stomach lesion segmentation image. According to the technical scheme of the invention, the gradual disappearance of the gradient caused by the deepening of the network is avoided, and the segmentation accuracy and precision are improved.

Description

technical field [0001] The invention relates to the technical field of image segmentation, in particular to a method for segmenting gastric lesions based on a generative confrontation network and a system for segmenting gastric lesions based on a generative confrontation network. Background technique [0002] Gastric cancer is a serious and fatal malignant tumor. Gastric cancer is found in the top five cancers with the highest incidence and the top five deaths in the world. At present, the commonly used examination method for gastric cancer is gastroscopy. Doctors can directly observe the internal conditions of the stomach through gastroscope pictures, so as to judge early gastric cancer. However, manually labeling the lesion area is a time-consuming and laborious task. Using computer-aided diagnosis technology to segment gastric lesions has become an effective way. Extracting features such as texture and color in gastroscopy images is an important means of identifying lesi...

AI Technical Summary

Problems solved by technology

In the identification of lesion areas in gastroscopy images, the results of identifying and segmenting gastric lesions through simple deep convolutional neural networks are not ideal

Images