A Dual-Model AI Framework for Alzheimer's Disease Diagnosis Using Clinical and MRI Data

Fatih Ciftci^1*Kadriye Yasemin Usta Ayanoglu^2*Sajjad Nematzadeh^3*Ferzat Anka¹

• ¹Fatih Sultan Mehmet Vakif Universitesi, Istanbul, Türkiye
• ²Technische Universitat Darmstadt, Darmstadt, Germany
• ³Istanbul Topkapi Universitesi, Istanbul, Türkiye

Alzheimer's disease (AD) is a progressive neurodegenerative disorder that demands advanced diagnostic strategies for early and accurate detection. This study presents a hybrid AI-driven diagnostic framework that synergistically integrates an Artificial Neural Network (ANN) and a Convolutional Neural Network (CNN) to improve AD classification. The ANN, trained on clinical data from 1,200 patients using 31 demographic, symptomatic, and behavioral features, provides an initial risk assessment, while the CNN processes 4,876 MRI images to classify AD into four stages with high precision. Our results demonstrate the ANN's accuracy of 87.08% in risk prediction and the CNN's superior performance of 97% in disease staging, with Grad-CAM visualizations enhancing interpretability. Unlike traditional diagnostic methods that often lack sensitivity and scalability, this dual-model approach effectively combines structured clinical data with imaging-based analysis, offering a more comprehensive and reliable assessment of AD. The integration of ANN and CNN not only strengthens diagnostic accuracy but also paves the way for AI-assisted clinical decision-making. Future work will explore lightweight CNN architectures and wearable technology for broader accessibility and early intervention.