Colorectal cancer digital pathological image discrimination method and system based on weak supervised learning

Abstract

The invention discloses a colorectal cancer digital pathological image discrimination method and system based on weak supervised learning. The colorectal cancer digital pathological image discrimination system based on weak supervised learning comprises an acquisition module used for collecting a colorectal cancer digital pathological image data set; a preprocessing module used for preprocessing data in the collected data set to obtain preprocessed data; a first classification module used for constructing a sampling block discrimination model based on a weak supervised learning algorithm, inputting the preprocessed data into the constructed sampling block discrimination model for processing, and obtaining a classification result of all pathological image blocks in a full-slice sampling packet; and a second classification module used for constructing a decision fusion model, inputting the obtained classification result of the pathological image blocks into the decision fusion model for fusion processing, and obtaining a classification result of the all-digital pathological image.

Description

technical field [0001] The invention relates to the technical field of pathological images, in particular to a colorectal cancer digital pathological image discrimination method and system based on weakly supervised learning. Background technique [0002] Tumor has always been one of the major diseases that endanger human health. Its morbidity and mortality rate are among the highest in the world, and it is still showing an upward trend year by year in developing countries. Colorectal cancer (CRC) is one of the most common gastrointestinal tumors. The 2018 China Cancer Statistics Report shows that the incidence and mortality of colorectal cancer in my country rank third and fifth among all malignant tumors, respectively. , among which the urban area is much higher than the rural area, and the incidence of colorectal cancer has increased significantly. Studies have shown that its incidence is related to the level of regional economic development, and is greatly affected by li...

AI Technical Summary

Problems solved by technology

[0006] Conventional deep learning algorithms require one-to-one labeling of training data to achieve good performance, and the quality of the data set will greatly affect the algorithm results

Due to the complexity of the digital pathology image format, the task of data labeling at the target magnification becomes difficult, and pathologists spend time and effort on data labeling. At the same time, there may be missing labels and errors in such a large-scale data. Standard and other issues

The way of finely labeling data obviously has certain limitations in the field of medical images

In addition, even with pixel-level labeled data, there is a problem of poor generalization between different data.

Images