Integrated Multi-Modal 3D-CNN and RNN Approach with Transfer Learning for Early Detection of Alzheimer’s Disease

Alzheimer’s disease (AD) is often first noticed through mild cognitive impairment (MCI), which involves minor variations in brain changes during intermediate stages. Although several research studies have focused on diagnosing AD in its primary stages, variations are observed in the complexity of brain and the functional magnetic resonance imaging (fMRI) used to study them make early detection of AD challenging. This study aims to advance the early diagnosis and monitoring of alzheimer’s disease by integrating structural MRI and functional fMRI data using a novel hybrid deep learning model. Leveraging a combination of 3D Convolutional Neural Networks (3D-CNN) and Recurrent Neural Networks (RNN), along with transfer learning techniques, we propose a robust framework for distinguishing between No Cognitive Impairment (NCI), Mild Cognitive Impairment (MCI), Subjective Cognitive Impairment (SCI), and AD. Using the publicly available ADNI dataset, which includes 694 structural MRI scans and 306 fMRI scans, we employed a five-fold cross-validation strategy, achieving a mean testing accuracy of 99.5%. The results demonstrate significant improvements in classification accuracy compared to existing methods, with the proposed model outperforming state-of-the-art techniques in terms of accuracy, precision, and F1 score. Detailed statistical analysis and error examination reveal the robustness and reliability of our approach. Furthermore, we discuss the potential clinical implications of our findings, emphasizing the model’s applicability in early diagnosis and treatment planning for AD. This study not only sets a new benchmark for AD classification but also highlights the importance of integrating multimodal imaging data for improved diagnostic accuracy. Future research directions include incorporating additional data types, performing longitudinal analyses, enhancing model interpretability, and developing user-friendly tools for clinical application.