An immunohistochemical quantitative analysis method based on YOLOv8 and mutual information self-supervised learning

Abstract

This invention provides an immunohistochemical quantitative analysis method based on YOLOv8 and mutual information self-supervised learning, comprising: constructing a first dataset by combining Ki-67 and PD-L1, and using the first dataset to train a YOLOv8 model to obtain a YOLOv8 pre-trained model; using the YOLOv8 pre-trained model to perform cell detection and sectioning on pathological images to obtain first analysis data; constructing a base dataset of Ki-67 and PD-L1 to obtain a second dataset; training a mutual information self-supervised learning model on the second dataset in conjunction with the Ki-67 and PD-L1 test dataset to obtain a trained mutual information self-supervised learning model; and using the trained mutual information self-supervised learning model to determine the expression state and perform quantitative analysis on the first analysis data to obtain immunohistochemical quantitative analysis data. This invention uses YOLOv8 and mutual information self-supervised learning to achieve efficient and automated analysis of pathological images, reducing the influence of human factors and improving the accuracy of immunohistochemical quantitative analysis.

Images