Rapid diagnosis and scoring method for full-scale pathological slices based on deep learning

Abstract

The invention relates to a method for rapid diagnosis and scoring of full-scale pathological slices based on deep learning. Preprocess the full-scale stained images of pathological sections; change the number of nodes in the fully connected layer and output layer of the traditional AlexNet neural network to meet the needs of practical problems, and select the marked training sample set to train two for diagnosis and scored AlexNet neural network model to extract the high-dimensional feature information of the lesion area; use the two improved AlexNet neural network models that have been trained to realize the diagnosis and scoring of full-scale pathological section staining maps; draw the probability according to the predicted probability of diagnosis The heat map visually identifies the lesion area, and at the same time, scores the lesion level of the tissue through the statistics of the proportion of the number of sampled small blocks of different lesion levels. The method of the invention can fully automatically realize the diagnosis and Gleason scoring of the full-scale prostate tissue pathological slices, and the accuracy rate and operation speed greatly exceed the average level of manual diagnosis.

Description

technical field [0001] The invention relates to the fields of image processing and medicine, in particular to a method for rapid diagnosis and scoring of full-scale pathological slices based on deep learning. Background technique [0002] In my country, malignant tumors have long been the leading cause of death for urban and rural residents, and the mortality rate of malignant tumors is at the highest level in the world, and it shows a continuous growth trend. Prostate cancer is the most common malignant tumor in the male reproductive system, and its morbidity and mortality are second only to lung cancer, ranking second in cancer deaths. At present, H&E stained pathological sections are still the gold standard for diagnosing tissue cancer. However, pathological diagnosis often relies on the subjective judgment of pathologists, which is prone to misdiagnosis and missed diagnosis. At the same time, due to the uneven professional level of pathologists and the uneven geographi...

AI Technical Summary

Problems solved by technology

[0003] 1. At present, most of the pathological diagnosis of prostate tissue is just a binary classification, that is, judging whether it is normal or cancer tissue, and rarely scoring the severity of prostate cancer;

[0004] 2. Traditional cancer detection mainly uses local detection of cell nuclei, and seldom focuses on staining images of full-scale pathological sections;

[0005] 3. Using the method of texture analysis plus classifier, features need to be manually extracted, and the accuracy rate is only about 85%;

[0006] 4. The calculation efficiency is low, and the diagnosis takes a long time

Images