Cancer pathological image automatic detection method and system

Abstract

The embodiment of the invention provides a cancer pathological image automatic detection method and system. The method comprises the following steps: extracting a foreground image set of a full-slice image set by adopting a preset extraction algorithm; performing block processing on the foreground image set to obtain a block image set, and extracting category labels of the labeled data set; inputting the block image set and the category labels into an EM model semi-supervised learning framework for model training to obtain a labeled data set distribution probability and an unlabeled data set distribution probability; calculating the distribution probability of the labeled data set and the distribution probability of the unlabeled data set according to a probability graph standardization algorithm to obtain a cancer probability distribution graph; and processing the full-slice image set based on the cancer probability distribution diagram to obtain a standardized automatic detection result. According to the method and system, a semi-supervised algorithm framework based on expectation maximization is provided for automatic cancer region detection of a full-slice pathological image, and accurate pathological typing and grading can be carried out on cancer primary region tissues.

Description

technical field [0001] The invention relates to the technical field of pathological image diagnosis, in particular to an automatic detection method and system for cancer pathological images. Background technique [0002] At present, the diagnosis of cancer still relies on manual diagnostic methods, especially breast cancer diagnosis. In order to obtain the gold standard of breast cancer diagnosis, breast histopathological diagnosis is usually required. Pathologists need to analyze the tissue sections, carefully observe the breast lymph node sections, carry out the pathological staging of the breast, and also need to observe the tissue morphology to determine the pathological grade. [0003] Manual diagnosis is prone to missed and false detections. Many related studies have found that for the same patient, different pathologists usually give different diagnostic results. Studies have pointed out that the diagnostic consistency of pathologists for breast cancer is only 75.3%....

AI Technical Summary

Problems solved by technology

Sliced ​​pathological images usually cannot be directly used as the input of the artificial intelligence network. Existing algorithms directly crop it as the standard input of the network, resulting in the loss of global information of the pathological image, resulting in the inability to extract the global features of the pathological image.

Or the image is down-sampled to obtain the standard input of the network, resulting in a decrease in the resolution of the pathological image, resulting in the inability to extract the features of local details

In the case of breast cancer pathological images, the former cannot accurately distinguish carcinoma in situ from invasive carcinoma; the latter affects the extraction of key features of mitosis

Images