Detection and Retinal Pathology Analysis of Early to Intermediate AMD and Chronic DME in OCT Scans Using Convolutional Neural Networks and You Only Look Once

Abstract

Purpose: Diabetes can lead to complications such as diabetic nephropathy, diabetic neuropathy, and diabetic macular edema (DME). Like DME, AMD-related vision loss can also be prevented through early detection and prompt treatment. Optical coherence tomography (OCT) is the most widely used imaging technique in ophthalmology for diagnosing these conditions. While OCT is effective for early screening, the increasing volume of OCT scans adds to the workload of ophthalmologists, who must interpret each image.
Method: To address this challenge, automated diagnostic screening systems are being actively developed to reduce the burden on eye care professionals. Prolonged diabetic may lead to Macular Edema (DME) or CNV or DRUSEN. Which causes the aberrant development of blood vessels from the choroid layer into the retina. DME, CNV may cause to vision loss. Early detection of the disease and correct treatment are crucial to address the issue of vision loss. Artificial intelligence techniques have been widely adopted in many research work to overcome these limitations. This paper contributes towards development of textual reports, helpful for ophthalmologists.
Result and Conclusion: This is the review developed for detecting and analysing DME, DRUSEN CNV using AI techniques over the last decade. Also, we tried to convert the OCT finding into a textual report, very well assessed by ophthalmologist. The system automates the process of analysing OCT images, eliminating the need for manual interpretation by ophthalmologists. This significantly reduces the time required for diagnosis. The system processes the images and generates textual reports using deep learning methods.

Keywords