Gastric early cancer histological image classification system based on deep neural network

Abstract

The invention relates to a gastric early cancer histological image classification system based on a deep neural network, and the system comprises a WSI pathological section obtaining module which is used for obtaining a gastric early cancer pathological section image; an image preprocessing module which is used for preprocessing the early gastric cancer pathological section image; a data set division module which is used for dividing the data set after image preprocessing into a training set and a verification set, a training set which is used for model training, and the verification set is used for verifying a model effect; a feature extraction module which is used for extracting picture features of the training set or the verification set based on a GoogleNet double-branch network structure and outputting feature vectors of corresponding pictures; and a classifier training module which is used for training based on the neural network model to obtain a gastric early cancer classification model.

Description

technical field [0001] The invention relates to a histological image classification system for early gastric cancer based on a deep neural network, and relates to the technical field of image classification. Background technique [0002] Early gastric cancer refers to gastric cancer whose lesions only invade the mucosa or submucosa. According to statistics, there are nearly 1 million cases of gastric adenocarcinoma every year worldwide, ranking third among the causes of cancer death in the world, and it is also the main cause of infection complicated by cancer death. Among them, early gastric cancer accounts for 15% to 57% of gastric cancer cases, so early diagnosis and effective treatment of early gastric cancer are of great significance to reduce the mortality of gastric cancer. Histocytological characteristics are currently recognized and the most reliable diagnostic criteria for tumors. Usually, the process of pathological diagnosis of early gastric cancer is to first ...

AI Technical Summary

Problems solved by technology

In order to adapt to the input of the neural network, there are two processing methods: image scaling and cropping. The former loses the local details of the image, and the latter lacks the global feature information of the original image, which can easily lead to over-fitting of the neural network.

The distribution between the processed data set samples is uncertain and likely to be unbalanced. If all samples are learned uniformly without distinction, it will easily lead to insufficient learning of the deep neural network under the limited data set, which will lead to different results. Inaccurate

Images