Alzheimer's disease classification by supervised and intelligent techniques

Area of Science:

• Neuroimaging and Machine Learning
• Biomedical Data Analysis

Background:

• Recent neuroimaging advancements, including PET and MRI, aid Alzheimer's disease (AD) and mild cognitive impairment (MCI) diagnosis.
• Combining imaging modalities with machine learning (ML) improves diagnostic accuracy for neurodegenerative diseases.

Purpose of the Study:

• Develop predictive models using pre-treatment brain imaging data.
• Accurately distinguish between normal controls (NC), MCI, and AD stages.
• Enhance diagnostic precision for Alzheimer's disease progression.

Main Methods:

• Utilized the Alzheimer's Disease Neuroimaging Initiative (ADNI) database.
• Processed 3D MRI, PET Florbetaben, and PET Flortaucipir images.
• Applied machine learning techniques: fuzzy logic, CNN, and MLP with features like amyloid-β, tau, and empty space volumes.

Main Results:

• Fuzzy logic achieved 99.1% classification accuracy, surpassing CNN (90.67%) and MLP (94%).
• Multimodal data integration significantly improved performance over single-modality approaches.
• Demonstrated superior performance of fuzzy logic in classifying AD stages.

Conclusions:

• Integrating advanced ML with multimodal neuroimaging effectively classifies Alzheimer's disease stages.
• Findings address critical gaps in early AD and MCI detection.
• Provides a foundation for future clinical applications in neurodegenerative disease diagnosis.