Digital pathological image classification method and system based on superpixel segmentation and image convolution

Abstract

The invention relates to a digital pathological image classification method and system based on superpixel segmentation and image convolution, and the method comprises the following steps: S1, carrying out superpixel segmentation of digital pathological images through a simple linear iterative clustering algorithm SLIC, and obtaining a superpixel segmentation region; S2, taking each super-pixel region as a node in a graph structure, taking whether the super-pixel regions share the same edge as the basis of edge distribution between the nodes, and constructing the graph structure based on the digital pathological images; S3, randomly dividing the constructed data of the graph structure into a training set, a verification set and a test set according to a preset proportion, and obtaining an optimal prediction model by training, verifying and testing a graph convolutional neural network; and S4, based on the prediction model, classifying the digital pathological images. According to the invention, classification prediction of the digital pathological images is realized by using the graph convolutional neural network, and the accuracy of pathological image classification is improved.

Description

technical field [0001] The invention relates to the technical field of computer-aided diagnosis, in particular to a method and system for classifying digital pathological images based on superpixel segmentation and graph convolution. Background technique [0002] A standardized and unified pathology report provides sufficient information for clinicians, but the morphological evaluation of pathological indicators increases the workload of pathologists. Due to the maturity and standardization of pathological biopsy techniques, the number of pathological biopsy samples for diagnosis continues to rise, and the diagnostic results will be affected by factors such as fatigue and subjective experience. In recent years, the rapid development of digital pathological scanning technology and artificial intelligence (AI) has provided great potential for pathological diagnosis. Specifically, a deep convolutional neural network (CNN) is used to extract high-dimensional features in patholo...

AI Technical Summary

Problems solved by technology

However, the traditional CNN processes structured two-dimensional array images with pixel values. This structured two-dimensional array data is difficult to express the relationship and histological characteristics between cells and glands in histopathology. Neglecting the spatial relationship between microscopic cells will lose some key feature information to improve model performance

Images