A Multimodal Deep Learning Framework Using ResNet-101 and Firefly-Based Feature Selection for Early Diagnosis of Dementia and Alzheimer’s Disease

Dementia and Alzheimer’s Diseases (AD) has global health challenges especially due to the progressive nature and the impact these diseases put on elderly population. Early detection is vital to improve prognosis and initiate timely intervention so that lives can be saved. This paper proposes a novel multimodal diagnostic framework that integrates clinical data and MRI imaging for the early classification of dementia and AD. The clinical pathway employs the OASIS dataset, wherein numerical features undergo statistical filtering followed by metaheuristic feature selection using the Firefly Algorithm. For imaging-based diagnosis, a deep Convolutional Neural Network (CNN) architecture is developed using MRI data, further optimized via a hybrid ResNet101–Support Vector Machine (SVM) model to enhance classification accuracy. The decision-level fusion module integrates outputs from both modalities to mitigate inter-class confusion and enhance diagnostic precision. When benchmarked against state-of-the-art models, the proposed framework achieves superior performance with an Accuracy of 0.933 and F1-Score of 0.921, highlighting its effectiveness and reliability. The results underscore the significance of integrated optimization and hybrid decision logic in advancing early-stage Alzheimer’s diagnosis.