Efficient Integration of Lightweight Semantic Segmentation for Retinal Fluid Analysis on Edge Devices

Our research merges electrical engineering with ophthalmology, creating a semantic segmentation model for retinal fluid analysis, optimized for edge devices. Enhancing CFPNet with lightweight structures and attention mechanisms, the model is efficient and accurate, trained on annotated retinal Optical Coherence Tomography (OCT) Images. With less than a million parameters, it achieves high Dice coefficients in lesion segmentation, improving medical imaging portability and utility, especially in remote areas, and offering potential advances in clinical image processing and engineering fields.