An Optimized Deep Learning Network for Prognosis of Alzheimer's Disease using Structural Magnetic Resonance Imaging

For the early diagnosis and treatment of Alzheimer's Disease (AD), accurate and early prediction is critical. To reliably predict patients with early stages of AD, i.e., Mild Cognitive Impairment (MCI), must be diagnosed to take preventive measures. Magnetic Resonance Imaging (MRI) can be used for the clinical non-invasive examination of patients with suspected AD. In the paper, an optimized deep learning network (DLN) model is proposed to automatically diagnose AD, MCI, and cognitive normal (CN). Hyperparameters play an important role in DLN's training and performance, so the hyperparameters' selection is the main step while designing the model. Therefore, the hyperparameters are optimized using a whale optimization algorithm (WOA) for training the layers of DLN. The proposed optimized DLN is tested on an openly accessible available dataset, the ADNI dataset, consisting of structural MRI (sMRI) images of AD, MCI, and CN. Performance of the optimized DLN is compared with the state of the art networks to prove the efficacy of the proposed model. The proposed DLN model can assist in the automatic screening of AD patients and decrease the burden of medicinal services frameworks.