Alzheimer's Disease Brain Areas: The Machine Learning Support for Blind Localization

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Machine Learning

Background:

• Analyzing Positron Emission Tomography (PET) scans for Alzheimer's disease (AD) diagnosis is challenging due to image noise, low resolution, and subtle differences between healthy and impaired cognition.
• High-dimensional classification methods are crucial for automatically discriminating between normal controls (NC), mild cognitive impairment (MCI), and individuals progressing to Alzheimer's disease (MCIAD).

Purpose of the Study:

• To develop and validate a voxel-based method for volumetric PET image analysis to accurately classify neurodegenerative conditions.
• To identify and visualize informative brain regions indicative of metabolic changes associated with Alzheimer's disease.

Main Methods:

• A voxel-based method for volumetric PET image analysis was developed.
• Three classification experiments were conducted: Alzheimer's disease vs. Normal Control, Alzheimer's disease vs. Mild Cognitive Impairment, and MCI converting to AD vs. MCI.
• Informative brain regions were identified using statistical features and visualized through 'maps' compared with voxel-wise statistics.

Main Results:

• The developed method achieved a classification rate of 95.5% using mean intensity from selected brain patches as classifier input.
• The method effectively extracts information about the location of metabolic changes related to Alzheimer's disease.
• Visualized 'maps' highlighted the most informative brain regions, correlating with voxel-wise statistics.

Conclusions:

• The proposed voxel-based PET image analysis method demonstrates high accuracy in classifying Alzheimer's disease and related cognitive impairments.
• This approach aids in identifying subtle metabolic changes and pinpointing key brain regions affected by Alzheimer's disease.
• The method shows potential for automated diagnosis and understanding disease progression in neurodegenerative disorders.