Application of Generalized Split Linearized Bregman Iteration algorithm for Alzheimer's disease prediction.

Area of Science:

• Neuroimaging
• Machine Learning
• Medical Diagnostics

Background:

• Alzheimer's disease (AD) diagnosis relies on clinical assessments and neuroimaging.
• Accurate and early detection of AD is crucial for effective patient management.
• Structural magnetic resonance imaging (sMRI) provides valuable data for AD detection.

Purpose of the Study:

• To develop and validate a novel machine learning algorithm for Alzheimer's disease detection using sMRI data.
• To assess the accuracy and generalization capability of the proposed classification method.
• To identify key brain regions predictive of Alzheimer's disease.

Main Methods:

• Application of a novel machine learning algorithm, Generalized Split Linearized Bregman Iteration (GSplit LBI), combining logistic regression and structural sparsity.
• Utilized a dataset of 57 Alzheimer's disease patients and 47 normal controls, extracting gray matter volume.
• Employed 10-fold cross-validation and cross-testing on independent datasets (ADNI, Chinese cohort) for validation.

Main Results:

• The GSplit LBI model achieved a classification accuracy of 90.44% in the local cohort.
• Identified top predictive voxels (6% of parameters) contributing to accurate AD classification.
• Demonstrated good performance and stable feature selection across different cohorts, including ADNI.

Conclusions:

• The developed GSplit LBI algorithm offers a highly accurate and reliable method for Alzheimer's disease detection from sMRI.
• The algorithm exhibits strong generalization capabilities, performing well on diverse patient populations.
• GSplit LBI shows potential for clinical application in early and accurate Alzheimer's disease diagnosis.