Skill Adam Optimization based Deep Spiking Neural Network for Retinal Multi Disease Classification in Fundus Images

The aim of fundus imaging is to inspect the irregularities corresponding to diseases that affect eye. The fundus image has a crucial role in detection and observation of diverse ophthalmic diseases. Existing researches have engaged on detection of single disease from the fundus image. However, simultaneous recognition of multi-disease (MD) from the fundus images is still confronting a major challenge. Therefore, this challenge can be addressed by devising a model to detect MD from the fundus imaging with an enhanced accuracy. Here, Skill Adam Optimization based Deep Spiking neural network (SAO_Deep SNN) is introduced for retinal MD classification using fundus images. Initially, considered fundus image is pre-processed to eliminate noises utilizing Median filter (MF). Thereafter, optic disc (OD) is segmented by DeepJoint segmentation and next, artery and vein (A/V) classification is accomplished using sparking process. Then, features are extracted from OD segmented image and A/V classified image in feature extraction (FE) phase. Finally, retinal MD classification is done by Deep Spiking Neural Network (Deep SNN). However, Deep SNN is trained utilizing Skill Adam Optimization (SAO) that is introduced by integrating Skill Optimization Algorithm (SOA) with Adam algorithm. Additionally, SAO_Deep SNN achieved accuracy of 90.81%, sensitivity of 91.69% and specificity of 90.49%.