Abstract: Detection of early age-related macular degeneration (AMD) is greatly time-consuming to effectively treat this condition. This study has taken an initiative to observe how Convolutional Neural Networks can detect early AMD signs more accurately by analyzing OCT images. This research was funded to determine whether a CNN would be able to identify incipient signs of AMD in OCT images, so we were working under the assumption of being able to predict some symptoms of the eye ailment with an accuracy rate of the order of nearly 100%, by a CNN model which had learned specific characteristics about the symptoms of AMD such as Choroidal Neovascularization, Diabetic Macular Edema, Drusen buildup through the OCT images. CNV is a process whereby new blood vessels grow in the choroid layer and leak, leading to severe vision loss. DME is one of the complications of diabetes, where fluid accumulates in the macula and causes it to swell; as a result, vision becomes blurred. Drusen are yellow deposits underneath the retina, usually indicative of AMD, which leads to vision deterioration. Trained on 83,416 images labeled with each symptom of AMD, the CNN model achieved a prediction accuracy of 93.75%. This accuracy is potentially high enough to render the model reliable in detecting macular degeneration based on certain OCT image features. Therefore, the study highlights that deep learning-driven analysis of OCT images has the potential to radicalize deformation in detecting and managing early AMD. In this respect, the integration of the CNN-based analysis into clinical practice may enhance the speed and accuracy of diagnosis, hence timely and effective therapeutic strategies against AMD.
Keywords: Age-related macular degeneration (AMD), Choroidal neovascularization (CNV), Convolutional Neural Network (CNN), Diabetic macular edema (DME), Deep learning, Drusen, Macula, OCT images, Prediction accuracy, Retinal analysis, Vision loss
|
DOI:
10.17148/IJARCCE.2024.13916
