Alzheimer disease predicting from clinical and MRI data using DeepALZNET dual pathway framework.

Area of Science:

• Neurology
• Computational Neuroscience
• Medical Imaging Analysis

Background:

• Alzheimer's Disease (AD) presents progressive cognitive decline, with diagnosis often delayed due to subtle early symptoms.
• The lack of effective cures emphasizes the critical need for early and accurate diagnostic tools to manage disease progression.
• Current advanced methods often require extensive data and computational resources, limiting practical application.

Purpose of the Study:

• To introduce DeepALZNET, a dual-pathway computational framework for enhanced Alzheimer's Disease prediction.
• To provide a practical, interpretable, and adaptable solution for early AD diagnosis using either clinical data or brain MRI.
• To develop a system that bridges the gap between research benchmarks and real-world deployable solutions.

Main Methods:

• A dual-pathway framework: one pathway for structured clinical data (1D CNN + Random Forest) and another for unstructured brain MRI scans (VGG19 transfer learning).
• Empirical validation on public datasets (2k clinical cases, 40k MRI images) and established benchmarks (ADNI, OASIS).
• Focus on practical applicability, interpretability, and adaptability, contrasting with resource-intensive transformer or attention-based methods.

Main Results:

• Both DeepALZNET pathways achieved competitive accuracy in predicting Alzheimer's Disease.
• The framework demonstrated robustness across different datasets and benchmarks.
• The system operates effectively using clinical data or MRI scans independently, highlighting its flexibility.

Conclusions:

• DeepALZNET offers a viable and practical computational framework for early Alzheimer's Disease detection.
• The dual-pathway design enhances diagnostic accuracy and applicability in clinical settings.
• Future work can explore multimodal fusion and attention mechanisms within the DeepALZNET framework.