Detecting pigment epithelial detachment (PED) in the Optical Coherence Tomography (OCT) images is crucial for diagnosing and treating retinal illnesses including age-related macular degeneration (AMD). Machine learning approaches, particularly deep learning algorithms, are becoming more popular for automated PED detection due to their capacity to identify complex patterns from OCT data. In this manuscript, Optimized Contrastive Multi-Level Graph Neural Networks based Pigment Epithelial Detachment Detection in OCT images (PEDD-OCTI-CMLGNN) is proposed. Here, the input image is gathered from Retinal OCT image dataset. To execute, image is pre-processed utilizing High Accuracy Distributed Kalman Filtering (HADKF) to eliminate the speckle noise. The pre-processed image is provided into Simple Contrastive Graph Clustering (SCGC) for eye’s RPE layer in OCT images. Then, feature extraction is done by Component Separable Synchroextracting Transform (CSST). Finally, the extracted features are given to Contrastive Multi-Level Graph Neural Networks (CMLGNN) to classify the Pigment Epithelial Detachment Detection as Normal and Abnormal. Multi Objective Generalized Normal Distribution Optimization (MOGNDO) is employed to optimize the weight parameter of CMLGNN. The PEDD-OCTI-CMLGNN is implemented and its efficiency is evaluated utilizing some performance metrics. The proposed PEDD-OCTI-CMLGNN method provides higher Accuracy; lower error rate and higher Sensitivity compared with existing methods
Nagarani et al. (Fri,) studied this question.