Title: Optimisation enabled deep learning for retinal multi-disease classification

Authors: Natarajan Mohana Suganthi; Mailerumperumal Arun

Addresses: Department of Computer Science and Engineering, Vel Tech Rangarajan Dr. Sagunthala R&D Institute of Science and Technology, Chennai, Tamil Nadu, 600062, India ' Department of Computer Science and Engineering, Vel Tech Rangarajan Dr. Sagunthala R&D Institute of Science and Technology, Chennai, Tamil Nadu, 600062, India

Abstract: The retina is a critical component of the human visual system, responsible for capturing light and transmitting visual signals to the brain. Various diseases can affect the retina, leading to impaired vision or even blindness. Hence, in this research, the remora puzzle optimisation-based hybrid deep learning model (RPOA_Hybrid DL) is developed for retinal disease classification. Initially, the pre-processing is done by Laplacian filter. The optic disc connects the retina to the optic nerve and is used to identify the illness. The Bayesian fuzzy clustering (BFC) is used for optic discs segmentation. The arteries and veins are classified using the retinal artery/vein classification network (AVNet). In feature extraction, the critical features are extracted. Furthermore, the RPOA-enabled EfficientNet-visual geometry group net (EfficientNet-VGG-16), termed RPOA_Hybrid DL is used to classify the retinal diseases. The optimal accuracy, sensitivity, and specificity offered by the model is 0.897, 0.884 and 0.894.

Keywords: Laplacian filter; remora optimisation algorithm; ROA; Bayesian fuzzy clustering; BFC; puzzle optimisation algorithm; POA; artery/vein classification network; AVNet.

DOI: 10.1504/IJBRA.2026.151288

International Journal of Bioinformatics Research and Applications, 2026 Vol.22 No.1, pp.20 - 49

Received: 08 May 2024
Accepted: 02 Sep 2024
Published online: 22 Jan 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article