Multifold Bayesian kernelization in Alzheimer's diagnosis.

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Accurate diagnosis of Alzheimer's Disease (AD) and Mild Cognitive Impairment (MCI) is critical for early dementia detection and treatment.
• Current diagnostic approaches often rely on machine learning models trained with multi-modal biomarkers, but accuracy is limited by model performance and data integration methods.

Purpose of the Study:

• To propose and validate a novel algorithm, Multifold Bayesian Kernelization (MBK), for improved diagnosis of AD and MCI.
• To model the diagnosis process as a synthesis analysis of multi-modal biomarkers, enhancing diagnostic accuracy.

Main Methods:

• MBK constructs biomarker-specific kernels maximizing local neighborhood affinity.
• A Bayesian framework evaluates individual biomarker contributions.
• A novel diagnosis scheme synthesizes individual biomarker probabilities for subject diagnosis.

Main Results:

• The MBK algorithm was validated using multi-modal neuroimaging data from the ADNI baseline cohort.
• The study included 85 AD, 169 MCI, and 77 cognitive normal subjects.
• MBK demonstrated significant diagnostic improvements across all groups compared to state-of-the-art methods.

Conclusions:

• MBK offers a superior approach to diagnosing AD and MCI by effectively synthesizing multi-modal biomarker data.
• The algorithm shows promise for enhancing early dementia detection and treatment planning.