An esophageal cancer pathological image processing method based on depth learning

Abstract

The invention discloses an esophageal cancer pathological image processing method based on depth learning, which comprises the following steps: a) pathological slice scanning; B) Encircling the epithelial region type, encircling the normal region of the epithelial region, the low-grade and high-grade precancerous lesion region; (c) image preprocess to obtain small epithelial images; (d) divide each epithelial small image into n image blocks along its longitudinal direction by a convolution neural network, and carry out feature extraction on each image block; E) a long-term and short-term memory network LSTM for acquiring feature vectors of small epithelial images; F) Classifier classification; G) Modeling and tuning, h) Accuracy calculation. The esophageal cancer pathological image processing method of the invention After processing by CNN, LSTM network and classifier, the probability of each epithelial small image being normal, low-grade and high-grade precancerous lesions is obtained, which provides an effective digital image processing method for the scientific utilization of pathological esophageal cancer slices. The method has remarkable beneficial effect and is suitable for application and popularization.

Description

technical field [0001] The present invention relates to a method for processing pathological images of esophageal cancer, and more specifically, to a method for processing pathological images of esophageal cancer based on deep learning (CNN+LSTM). Background technique [0002] Esophageal cancer (esophageal cancer, EC) is a malignant tumor of the digestive tract originating from the esophageal mucosal epithelium, and about 300,000 people worldwide die of esophageal cancer every year. my country is one of the countries with high incidence and high mortality rate of esophageal cancer in the world. According to the 2017 data of "China Tumor Registration Annual Report", esophageal cancer ranks fourth in the mortality rate of malignant tumors. At present, more than 90% of esophageal cancer patients in my country have progressed to the middle and late stage when they are diagnosed, the prognosis is poor, and the quality of life is low. The total 5-year survival rate of each stage o...

AI Technical Summary

Problems solved by technology

[0004] In terms of cancer diagnosis, deep learning has achieved certain results in the pathological diagnosis of skin cancer, breast cancer, gastric cancer, colon cancer, etc., and can detect abnormal lesions from X-rays, CT scans and MRI images. No progression in the pathological diagnosis of the lesion

The pathology department of the hospital has a large number of full pathological slides of esophageal cancer. These pathological slides of esophageal cancer form the pathological samples of esophageal cancer. These samples should be used to provide scienti

Images