Convolutional Neural Network-based MR Image Analysis for Alzheimer's Disease Classification.

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Medical Diagnostics

Background:

• Alzheimer's disease (AD) and mild cognitive impairment (MCI) are progressive neurological conditions.
• Accurate classification of AD, MCI, and normal controls (NC) is crucial for timely intervention.
• Magnetic resonance (MR) imaging offers insights into brain structure, particularly the hippocampus.

Purpose of the Study:

• To develop and evaluate a convolutional neural network (CNN) for classifying AD, MCI, and NC subjects.
• To assess the efficacy of using hippocampus region images for disease classification.
• To determine the diagnostic performance of the proposed CNN model.

Main Methods:

• Brain MR images from the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset were used.
• Hippocampal regions were automatically segmented using 3D-Slicer and the ICBM template.
• A CNN model was trained on preprocessed hippocampus images, with datasets grouped for binary classification (AD/NC, AD/MCI, MCI/NC).

Main Results:

• The CNN model achieved high classification accuracies: 92.3% for AD/NC, 85.6% for AD/MCI, and 78.1% for MCI/NC.
• Preprocessing involved local entropy minimization (LEMS) for inhomogeneity correction.
• The model demonstrated robust performance across different classification tasks.

Conclusions:

• The proposed CNN method demonstrates high accuracy in classifying Alzheimer's disease, mild cognitive impairment, and normal controls.
• Using small hippocampus image patches is an efficient strategy for disease classification.
• This approach shows promise as a valuable tool for neurodegenerative disease diagnosis.