ML-Driven Alzheimer's disease prediction: A deep ensemble modeling approach

Area of Science:

• Neurology
• Artificial Intelligence
• Medical Imaging

Background:

• Alzheimer's disease (AD) is a progressive neurological disorder causing cognitive decline.
• Early and accurate AD detection is crucial for management and treatment.
• Deep learning shows promise for analyzing medical images for diagnosis.

Purpose of the Study:

• To develop an ensemble learning framework for improved Alzheimer's disease diagnosis.
• To combine multiple deep learning architectures for enhanced diagnostic accuracy.
• To validate the model's performance on diverse datasets for clinical applicability.

Main Methods:

• An ensemble learning framework integrating VGG16, VGG19, ResNet50, InceptionV3, and EfficientNetB7.
• Training and testing on a dataset of 3,714 MRI brain scans from Iraq (NonDemented, MildDemented, VeryDemented).
• Validation on external datasets: OASIS and ADNI.

Main Results:

• The proposed voting ensemble model achieved 99.32% diagnostic accuracy on the primary dataset.
• External validation demonstrated high performance: 86.6% on OASIS and 99.5% on ADNI.
• The model exhibited high precision and recall across all dementia stages.

Conclusions:

• Ensemble learning effectively enhances the accuracy of Alzheimer's disease diagnosis using MRI scans.
• The developed model is a reliable and robust tool for early AD detection.
• The findings suggest significant potential for clinical application in neurodegenerative disease diagnosis.