MRI-Based Early Detection and Multi Stage Classification of Alzheimer’s Disease using a Hybrid CNN-LSTM Deep Learning Model

Abstract

The most prevalent cause of dementia, Alzheimer's disease (AD), is linked to a persistent deterioration in mental and cognitive capacities. Dementia refers to the gradual deterioration of neuropsychiatric functions, which affects memory, reasoning, and the ability to perform daily activities. As they show structural alterations in various brain regions, brain medical images especially Magnetic Resonance Imaging (MRI) are frequently used to assess the formative stages of Alzheimer's disease. Deep learning algorithms and sophisticated computer-based methods have been employed more frequently in recent years to identify and categorize Alzheimer's disease. In order to diagnose Alzheimer's disease early and classify it into multiple classes using MRI scans, this study uses a hybrid deep learning architecture that blends Deep Convolutional Neural Networks (DCNN) and Long Short-Term Memory (LSTM) networks. High-level spatial characteristics are extracted from brain pictures by the DCNN component, and to improve classification accuracy, the LSTM layer records sequential relationships within these features. To improve prediction on unseen data, class-weighted learning and regularization techniques are applied to address data imbalance and prevent overfitting. Standard performance criteria including accuracy, precision, recall, F1-score, and validation loss are used to assess the suggested CNN–LSTM model. According to experimental results, the suggested strategy outperforms traditional CNN-based techniques in properly identifying various stages of Alzheimer's disease, with a classification accuracy of 97%.