Diabetic retinopathy classification method based on multi-scale convolutional neural network

Abstract

The invention discloses a diabetic retinopathy classification method based on a multi-scale convolutional neural network, and the method comprises the steps: carrying out normalization preprocessing of an original fundus image, taking a target pixel as the center, carrying out multi-scale segmentation on the preprocessed image and obtaining a series of image sequences with the same shape and sizeunder different scales; for each scale, performing feature extraction on the image sequence by using a convolutional neural network model to obtain features of different scales, performing fusion, andperforming final classification on the fused features to obtain a lesion detection result of the image sequence (i.e., a target pixel); and integrating and outputting detection results of all pixel points on the original fundus image to obtain a diabetic retinopathy detection result image with lesion positioning and classification. According to the method, multi-scale feature extraction and fusion are carried out, the self-adaptive scale diabetic retinopathy of MAs and HEs can be detected at the same time, and the performance of a diabetic retinopathy detection algorithm is improved.

Description

technical field [0001] The present invention relates to medical image processing in computer vision, and more specifically, to a diabetic retinopathy classification method based on a multi-scale convolutional neural network. Background technique [0002] Medical image processing is one of the most extensive research topics in the field of computer vision in recent years. Diabetic retinopathy is one of the chronic blinding diseases and one of the main causes of preventable blindness in the world. The key to the treatment of this disease lies in early diagnosis and intervention treatment. Computer-aided diagnosis technology based on color fundus images can significantly improve the efficiency and effectiveness of early diabetic retinopathy screening. Therefore, it is necessary to develop a fast and reliable intelligent detection method for diabetic retinopathy based on this technology. [0003] The key to the detection of diabetic retinopathy is the location and classificatio...

AI Technical Summary

Problems solved by technology

Existing lesion-level diabetic retinopathy detection methods all have a common shortcoming: they only focus on a single type of lesion

This is because microvascular tumors and retinal hemorrhages have huge differences in shape and size, and the common detection method is to first divide the retinal image into a series of image sequences of fixed size and shape, and then perform lesion detection on the image sequences sequentially, so it is difficult to achieve Simultaneous and accurate detection of two types of lesions with different sizes and shapes

Images