Inherent Structure-Based Multiview Learning With Multitemplate Feature Representation for Alzheimer's Disease...

Area of Science:

• Neuroimaging
• Machine Learning
• Medical Diagnostics

Background:

• Multitemplate analysis of brain morphometrics aids Alzheimer's disease (AD) and mild cognitive impairment (MCI) diagnosis.
• Existing methods assume simple data distributions, which may not reflect real-world complexity.

Purpose of the Study:

• To develop an inherent structure-based multiview learning method for improved AD/MCI classification.
• To address limitations of current multitemplate approaches by not assuming pre-defined data distributions.

Main Methods:

• Extracting multiview feature representations using multiple templates.
• Clustering subjects within classes into subclasses and encoding them with unique codes.
• Employing a multitask feature selection model and an ensemble of support vector machine classifiers.

Main Results:

• The proposed method achieves promising results in AD/MCI classification.
• Demonstrated superior performance compared to existing state-of-the-art multitemplate-based methods.
• Validated on the Alzheimer's Disease Neuroimaging Initiative database.

Conclusions:

• The novel inherent structure-based multiview learning method enhances AD/MCI classification accuracy.
• This approach offers a more robust alternative to existing methods by accommodating complex data distributions.