Latent Space Learning and Feature Learning using Multi-template for Multi-classification of Alzheimer's Disease

Area of Science:

• Neurology
• Medical Imaging
• Artificial Intelligence

Background:

• Alzheimer's disease (AD) is a prevalent neurodegenerative disorder in aging populations, characterized by cognitive and behavioral impairments.
• The increasing prevalence of AD necessitates advanced diagnostic tools for early and accurate detection.
• Computer-aided diagnosis (CAD) systems are gaining traction for their potential to assist in AD assessment.

Purpose of the Study:

• To propose a novel computational model for multi-classification of Alzheimer's disease.
• To enhance diagnostic performance by integrating latent space learning and feature learning from multiple imaging templates.
• To identify and utilize the most discriminative features for improved AD classification.

Main Methods:

• A novel model combining latent space learning and feature learning was developed.
• Features were extracted from multiple templates to capture diverse aspects of brain data.
• Latent space learning identified inter-template relationships, while feature learning explored intrinsic feature correlations.
• Discriminative features were selected to optimize multi-classification performance.
• The Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset was utilized for model evaluation.

Main Results:

• The proposed model demonstrated competitive performance in multi-class AD classification.
• Comparative experiments validated the effectiveness of the combined latent and feature learning approach.
• The method successfully leveraged multi-template features for enhanced diagnostic accuracy.

Conclusions:

• The developed model offers a promising approach for computer-aided diagnosis of Alzheimer's disease.
• Integrating latent space and feature learning with multi-template analysis improves AD classification.
• This methodology holds potential for clinical application in early AD detection and management.