Method and terminal for processing DME typing based on deep neural network

Abstract

The invention provides a method and terminal for processing DME typing based on a deep neural network. The method comprises the steps: preprocessing a to-be-recognized OCT image; performing image feature extraction on the preprocessed OCT image through a trained DME feature extraction model, wherein the DME feature extraction model is obtained based on deep learning network training; processing the extracted image features to obtain a binary classification function value corresponding to whether a preset DME in the OCT image appears or not, wherein the preset DME comprises a DRT, a CME and anSRD; and obtaining a result of the binary classification task based on the binary classification function value and a preset threshold. According to the scheme, rapid and accurate identification of different types of DMEs is achieved.

Description

Technical field[0001]The present invention relates to the fields of clinical medicine ophthalmology and computer engineering, in particular to a method and terminal for processing DME typing based on a deep neural network.Background technique[0002]DME is the primary cause of vision loss in diabetic patients. According to its morphology in OCT examination, DME can be divided into diffuse retinal thickening (DRT), cystoidmacular edema (CME), and serous retina Separation (SRD, serous retinal detachment) and a mixed DME (Mixed DME) that mixes the above two or three types. On the OCT image, DRT showed an increase in the thickness of the retinal neuroepithelial layer, accompanied by a decrease in the inter-layer reflex of the neuroepithelial layer, and an expansion of the low-reflection area; Honeycomb, when the edema is obvious, the small cysts can fuse into larger cysts, and even only a thin inner limiting membrane remains in the fovea; SRD is the detachment of the neuroepithelial layer...

AI Technical Summary

Problems solved by technology

The main reason is that the traditional prediction method is highly subjective, there are many and complex factors to be considered comprehensively, and a large part depends on the ophthalmologist's clinical experience and knowledge level

And with the increasing prevalence of diabetes worldwide, it may bring a huge burden to the clinical diagnosis, treatment and management of DME patients

Therefore, for young doctors who lack clinical experience and doctors in community or primary hospitals with relatively low medical level, it is difficult to make an accurate DME classification

At present, there is still no fast, accurate, versatile, and widely used method for automatic and accurate classification of DME for clinical use.

Images