Exploring transfer learning techniques for classifying Alzheimer's disease with rs-fMRI.

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Neurology

Background:

• Alzheimer's disease (AD) is the most common cause of dementia, progressively impairing memory and cognitive function.
• Accurate diagnosis of AD remains challenging due to the complexity of brain structure and function.
• Resting-state functional magnetic resonance imaging (rs-fMRI) offers a non-invasive method to study brain activity in neurological diseases.

Purpose of the Study:

• To investigate the efficacy of deep learning algorithms in classifying Alzheimer's disease patients versus normal controls using rs-fMRI data.
• To compare the performance of VGG19, AlexNet, and ResNet50 deep learning models for AD diagnosis.
• To evaluate classification performance using accuracy, precision, recall, and F1-score.

Main Methods:

• rs-fMRI data from 97 participants (56 AD, 41 controls) were obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database.
• Data underwent extensive preprocessing before being used for classification.
• Transfer learning was employed with VGG19, AlexNet, and ResNet50 models for binary classification.

Main Results:

• Classification accuracies were 96.91% (VGG19), 98.71% (AlexNet), and 98.20% (ResNet50).
• AlexNet demonstrated superior performance across all evaluation metrics, including precision, recall, and F1-score.
• ResNet50 provided better interpretability via Grad-CAM visualizations, highlighting clinically relevant brain regions.

Conclusions:

• Deep learning models, particularly AlexNet, show high potential for accurate and automated diagnosis of Alzheimer's disease using rs-fMRI.
• The findings suggest that rs-fMRI combined with deep learning can serve as a valuable tool for AD detection and research.
• Further investigation into model interpretability, like with ResNet50, can enhance clinical translation and understanding of AD pathophysiology.