Joint Multi-Modal Longitudinal Regression Classification

Joint Multi-Modal Longitudinal Regression Classification performs simultaneous regression and classification on longitudinal multi-modal clinical data to predict Alzheimer's Disease outcomes and identify associated biomarkers using data from the Alzheimer's Disease Neuroimaging Initiative (ADNI).

Key Features:

• Multi-Modal Integration: Combines genetic information, brain imaging (brain scans), and other clinical modalities into a joint analysis of longitudinal data.
• Simultaneous Regression and Classification: Models continuous outcomes such as cognitive scores and categorical outcomes such as Alzheimer's Disease diagnosis within a unified framework.
• Regularization Techniques: Employs regularization methods to select and identify biomarkers relevant to Alzheimer's Disease.
• Iterative Algorithm with Convergence Proof: Formulates a non-smooth optimization problem and solves it using an efficient iterative algorithm that is proven to converge.
• Longitudinal Modeling: Explicitly models temporal trajectories across timepoints to leverage longitudinal information.

Scientific Applications:

• Alzheimer's Disease prediction: Applied to the ADNI cohort to predict cognitive status and clinical diagnosis of Alzheimer's Disease.
• Biomarker identification for AD: Identifies potential biomarkers associated with Alzheimer's Disease through regularization and multi-modal integration.
• Longitudinal progression analysis: Analyzes disease progression over time using longitudinal ADNI data.
• Extension to other biomedical studies: Applicable to other biomedical research involving longitudinal multi-modal clinical data.

Methodology:

Combines genetic data, brain imaging, and other clinical information for joint modeling of longitudinal data; formulates and solves a non-smooth optimization problem via an iterative algorithm with a convergence proof; applies regularization techniques for biomarker selection; validated by experiments on the ADNI cohort.