A sd-oct image retinopathy detection system based on class discrimination and localization

Abstract

The invention discloses a SD-OCT image retinopathy detection system based on category discrimination and positioning, which includes a data preprocessing module, a neural network prediction module, a heat map calculation module, a lesion area positioning module, and an accurate detection module. Perform enhancement processing, and then classify through the global convolutional neural network. According to the prediction results, obtain the heat map of the category mapping feature. Finally, by obtaining the larger activation value area, generate a mask to obtain the minimum circumscribed rectangle, and determine it according to the coordinates of the minimum circumscribed rectangle The location of the fundus lesion can be used to locate the fundus lesion area in the image. The present invention can effectively improve the accuracy of detection and recognition, and at the same time classify and recognize the retinal images of the fundus, it can locate the location of the corresponding fundus lesion according to the category discrimination information, and has important guidance for the subsequent diagnosis, analysis and treatment of the fundus significance.

Description

technical field [0001] The invention relates to the field of medical technology, in particular to a frequency-domain optical coherence tomography image retinopathy detection system based on category discrimination and positioning. Background technique [0002] The retina is located in the inner layer of the eyeball wall. It is a transparent film composed of the pigment epithelium and the retinal sensory layer. It is an extremely important structure in the eye. At the same time, the retina is very fragile. Many eye diseases occur inside the retina. Due to people's working life, eating habits and the aging of the population, the number of patients with fundus diseases is increasing year by year. Many patients have not received timely screening and effective treatment in the early stage, which leads to aggravation of the disease and even blindness. In recent years, with the development and maturity of Special-Domain Optical Coherence Tomography (SD-OCT), using the low coherence...

AI Technical Summary

Problems solved by technology

[0007] Aiming at the problems that the current algorithm has low detection accuracy of retinal lesions in a large number of SD-OCT images, the resolution of detection results is not high, and the cost of positioning identification is high, the present invention proposes a frequency-domain optical system based on a category discrimination positioning model. Coherent tomography image retinopathy detection system solves the problem of detecting and identifying a large number of fundus and locating and detecting the location of the lesion

Images