Hybrid Deep Learning Architecture for Accurate Classification of Alzheimer’s Disease using MRI Images

Alzheimer’s Disease (AD) is a progressive neurodegenerative disorder characterized by cognitive decline and memory loss. Precise and timely identification of this condition is significant for proper treatment. Magnetic Resonance Imaging (MRI) is a widely employed technology for analyzing brain structures and identifying abnormalities interconnected with AD. The recent advancements in medical fields, including the introduction of Deep Learning (DL) approaches, have automated the AD diagnosis process. However, achieving high accuracy remains a challenge due to the complexity and heterogeneity of AD manifestations. In this proposed work, we introduce a novel hybrid deep learning architecture designed to improve the accuracy of AD classification. The proposed architecture combines the strengths of convolutional neural networks (CNNs) for feature extraction and Autoencoder for capturing temporal dependencies in sequential MRI scans. Utilizing the Dragonfly optimization technique for fine-tuning the parameters and hyperparameters of deep learning models utilized in AD classification. The developed strategy was trained and validated using the publicly available AD and Munich databases. The proposed method achieved an overall accuracy of $98.34 \%$, precision of $99.06 \%$, recall of $99.65 \%$, and f-measure of $99.90 \%$. We anticipate that the proposed hybrid architecture will improve AD classification models’ discriminative power and robustness, thereby offering early and accurate diagnosis of this debilitating disease.