Applied Intelligence and Computing
Revolutionizing Retinal Health: AI-Driven Analysis for Early Pathology Detection in OCT Scans
Authors: Mohit Beri, Kanwarpartap Singh Gill, Deepak Upadhyay and Swati Devliyal
Publishing Date: 05-11-2024
ISBN: 978-81-955020-9-7
Abstract
The study's central focus is the critical need for early diagnosis of retinal abnormalities through the application of ground-breaking techniques in medical image processing. Despite the high volume of around 30 million OCT scans taken annually, the work proves that transfer learning is successful in medicine utilising a dataset of less than 1,000 retinal OCT photos. The dataset has been meticulously assembled utilising a hierarchical grading system that incorporates ophthalmologists, senior retinal specialists, undergraduates, and medical students. The four parts are as follows: DRUSEN, NORMAL, CNV, and DME. The method relies on a ResNet18 Convolutional Neural Network that has already been trained. The goal of retinal illness classification is being addressed by fine-tuning the last layer. The study's remarkable 94% accuracy rate shows that transfer learning has great potential as a powerful tool for early and precise detection of retinal disorders. Not only does this work add significantly to medical imaging, but it also provides a comprehensive reference for those just starting out in the field who are interested in using CNNs and TL.
Keywords
Artificial Intelligence, Deep Learning, Retinal Health Classification, Model Training, ResNet18 CNN Model.
Cite as
Mohit Beri, Kanwarpartap Singh Gill, Deepak Upadhyay and Swati Devliyal, "Revolutionizing Retinal Health: AI-Driven Analysis for Early Pathology Detection in OCT Scans", In: Mukesh Saraswat and Rajani Kumari (eds), Applied Intelligence and Computing, SCRS, India, 2024, pp. 65-71. https://doi.org/10.56155/978-81-955020-9-7-8