Alzheimer's disease diagnosis from diffusion tensor images using convolutional neural networks.

Area of Science:

• Neuroimaging
• Machine Learning
• Neurodegenerative Diseases

Background:

• Machine learning algorithms are increasingly used for neurodegenerative disease classification.
• Alzheimer's Disease (AD) and Mild Cognitive Impairment (MCI) detection remains a challenge.

Purpose of the Study:

• To develop a robust, low-cost classification system for AD and MCI against healthy controls (HC).
• To evaluate the efficacy of shallow convolutional neural networks (CNNs) using multimodal neuroimaging data.

Main Methods:

• Utilized the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset.
• Employed a convolutional neural network (CNN) classifier.
• Input data included diffusion maps (Diffusion Tensor Imaging - DTI) and gray-matter (GM) volumes (structural Magnetic Resonance Imaging - MRI).
• Adopted a ten-fold cross-validation scheme.

Main Results:

• Achieved an Area Under the Curve (AUC) of 0.94 for AD/HC classification and 0.84 for MCI/HC classification.
• Reported high accuracy, sensitivity, and specificity for both classification tasks.
• Demonstrated the effectiveness of combining structural MRI and DTI data for improved classification performance.

Conclusions:

• The developed CNN system provides a robust and cost-effective method for classifying AD and MCI.
• Incorporating multimodal imaging data significantly enhances classification accuracy.
• This approach shows promise for early detection and the development of interventions for Alzheimer's Disease.