Discrete migratory bird optimizer with deep transfer learning aided multi-retinal disease detection on fundus imaging

Retinal fundus diseases may cause permanent blindness without timely diagnosis and appropriate medical care. Single disease-based deep learning (DL) tactics were established for the diagnoses of age-related macular degeneration, glaucoma, and diabetic retinopathy. The upsurge in retinal diseases is alarming as it may cause irreversible visual loss if not treated early. Automation of diagnosis methods of retinal disease helps ophthalmologists save time and decision-making. Various research workers have addressed automated retinal disease classification; however, they are constrained to binary classification or handcrafted feature selection (FS). In recent years, DL-based techniques have been found useful for the automated classification of diseases using retinal photographs. This manuscript presents a Discrete Migratory Bird Optimizer with Transfer Learning Aided Multi-Retinal Disease Detection (DMBOTL-MRDD) method on Fundus Imaging. The main aim of the DMBOTL-MRDD method is to detect and classify multiple retinal diseases based on fundus photographs. In the DMBOTL-MRDD technique, the Wiener filtering (WF) approach is used for noise removal. In addition, ShuffleNetv2, a lightweight convolutional neural network (CNN) model, is used for feature vector extraction. Moreover, the DMBO technique is used for optimum hyperparameter selection of the ShuffleNetv2 model. Furthermore, a multi-layer autoencoder (ML-AE) is employed to detect multiple retinal diseases, including AMD, DR, and Glaucoma. The DMBOTL-MRDD technique is simulated under a benchmark fundus imaging dataset. The experimental validation of the DMBOTL-MRDD technique portrayed a superior accuracy value of 97.12% over existing techniques.