Multimodal neuroimaging-based deep learning framework for pattern analysis and early prediction of neurodegenerative diseases.
Authors
Affiliations (2)
Affiliations (2)
- Department of Electronics and Communication Engineering, Jyothi Engineering College, Thrissur, Kerala, India; APJ Abdul Kalam Technological University, Thiruvananthapuram, Kerala, India. Electronic address: [email protected].
- APJ Abdul Kalam Technological University, Thiruvananthapuram, Kerala, India; Department of Artificial Intelligence and Data Science, Jyothi Engineering College, Thrissur, Kerala, India. Electronic address: [email protected].
Abstract
Neurodegenerative diseases, such as Mild Cognitive Impairment (MCI) and Alzheimer's, pose significant challenges due to their progressive nature and late diagnosis. Early detection remains difficult, particularly when using conventional machine learning approaches that fail to capture complex spatial and temporal patterns in multimodal clinical data. Motivated by the need for accurate, scalable, and clinically applicable diagnostic tools, this study proposes a hybrid deep learning framework combining Convolutional Neural Networks (CNN) with Optimized Spatial-Temporal Bidirectional Gated LSTM (O‑SBGC‑LSTM). The framework is evaluated on 1,000 multimodal samples, achieving 94.8% accuracy, 93.9% precision, 94.2% recall, and a 94.0% F1-score, outperforming SVM (82.4%), Random Forest (85.7%), and CNN-LSTM (92.5%). Cross-validation confirms robustness (93.8-95.0% accuracy). The approach balances class performance across cognitively normal, MCI, and Alzheimer's cases. Future work will extend this framework to larger, multi-centre datasets and explore real-time clinical deployment, aiming to enhance early diagnosis, reduce misclassification, and support personalized treatment strategies for neurodegenerative disorders.