Pathological image cell nucleus detection method based on improved Faster RCNN algorithm

Abstract

The invention provides a pathological image cell nucleus detection method based on an improved Faster RCNN algorithm. The pathological image cell nucleus detection method comprises a tissue pathological section image feature extraction network; based on the deep learning theory, tissue pathological section images are processed through methods such as data enhancement transformation, difficult sample mining and small target detection optimization, and the target detection difficulty is reduced; and an over-fitting phenomenon is avoided according to cross validation, and regional nomination is completed by combining classification loss and positioning loss exploration in a loss function. According to the method, the detection speed of each image reaches 1s, the real-time requirement of practical application is met, good robustness is shown, and the method has the advantages of being high in detection precision and high in recall rate.

Description

technical field [0001] The invention relates to the field of electronic information, in particular to a pathological image nucleus detection method based on an improved Faster RCNN algorithm. Background technique [0002] The automatic detection and classification of cell nuclei and mitosis in pathological images is a hot research field today. At present, computer analysis algorithms for pathological images can be roughly divided into two categories: [0003] One is the traditional method of using manual design features plus machine learning classification, which relies on digital image processing technology or computer vision technology, and generally requires manual design features. For pathological image analysis, this requires domain expertise to define morphological and texture features that describe nuclei and mitotic figures. Most researchers use histogram of oriented gradient features (Histogram of Oriented Gradient, HoG), local binary pattern features (Local Binary...

AI Technical Summary

Problems solved by technology

However, this method processes the entire slice image pixel by pixel, and it takes several minutes to process a 4M slice image, and the efficiency is relatively low.

Images