Detection method for eyeground multi-disease classification based on deep learning

Abstract

The invention discloses a detection method for eyeground multi-disease classification based on deep learning. The method comprises the steps of obtaining an eyeground image, and selecting a training set and a test set according to independent identically-distributed conditions; preprocessing and enhancing the eyeground image, and extracting eyeground color photo image features; establishing a convolutional neural network model based on an EffecientNet + Attention Module, and adjusting the network structure; initializing network parameters by using a migration learning mode; based on the extracted eyeground features, training a detection model on a training set by using a reverse gradient propagation mode in deep learning, and finely adjusting parameters of the detection model; and finally,obtaining a detection model capable of distinguishing eight classifications including normal eyeground, high myopia, glaucoma, maculopathy, vein occlusion, diabetic retinopathy, hypertensive retinopathy and other degeneration. According to the detection method disclosed by the invention, automatic detection for multiple diseases of patients with eyeground diseases is realized, and the detection method is high in precision and high in speed.

Description

technical field [0001] The invention relates to a deep learning-based detection method for classification of multiple fundus diseases, belonging to the field of medical image processing. Background technique [0002] In recent years, the incidence of many fundus diseases has been high and the harm is serious. Fundus color photo examination is the most convenient and effective way to find fundus diseases. Doctors use professional fundus cameras to collect fundus images and then conduct manual diagnosis. Fundus examination requires professional fundus doctors and examination equipment. The investment in equipment is high, and it is difficult to train doctors. However, there are a large number of people in need in primary hospitals and physical examination centers, and the problem is very serious. Even if the number of instruments and doctors is sufficient, the diagnosis is very dependent on the personal status and experience of the doctors. Therefore, with the continuous dev...

AI Technical Summary

Problems solved by technology

In the existing multi-disease diagnosis methods, there are also shortcomings of complex models and low detection efficiency

Images