Enhanced Alzheimer's Disease Diagnosis Using Modified BrainNet3D: A Robust Framework for 3D MRI-Based Classification

Alzheimer's Disease (AD) is the leading cause of dementia globally, imposing significant societal and economic burdens. Early diagnosis, particularly in the prodromal stages, is crucial for effective management but remains challenging due to the subtle nature of structural brain changes and limitations in existing diagnostic frameworks, such as noise handling and computational scalability. This study proposes BrainNet3D, an optimized 3D convolutional neural network (CNN) model, to classify Alzheimer's stages-Cognitive Normal (CN), Mild Cognitive Impairment (MCI), and Alzheimer's Disease (AD)-using volumetric MRI data. The framework incorporates advanced preprocessing techniques, including Bayes Wavelet Transform for noise reduction and improved fuzzy c-means clustering for segmentation. Post-processing enhancements, such as gamma correction and Gaussian blur, further improve image clarity. The modified BrainNet3D architecture integrates dropout regularization, dynamic padding, and expanded convolutional blocks to enhance accuracy and scalability. Evaluated on the ADNI dataset, the model achieved a test accuracy of 96.05%, an average F1-score of 95.91 %, and robust performance across CN, MCI, and AD classifications. It outperformed state-of-the-art models, validating its effectiveness for early-stage AD diagnosis. BrainNet3D offers a scalable and automated diagnostic solution, supporting timely clinical interventions and reducing diagnostic burdens. Future directions include multimodal data integration, real-time preprocessing, and validation of diverse neurodegenerative diseases to expand their applicability in real-world clinical settings.