A LAG FUNCTIONAL LINEAR MODEL FOR PREDICTION OF MAGNETIZATION TRANSFER RATIO IN MULTIPLE SCLEROSIS LESIONS.

Area of Science:

• Statistics
• Medical Imaging
• Neuroscience

Background:

• Multiple Sclerosis (MS) causes myelin sheath damage in the central nervous system.
• Magnetic Resonance Imaging (MRI) measures tissue damage via the magnetization transfer ratio (MTR).
• Accurate MTR estimation in MS lesions is crucial for research and clinical applications.

Purpose of the Study:

• To propose a novel lag functional linear model for predicting responses using noisy, discrete functional data.
• To develop and compare two distinct procedures for estimating regression parameter functions within this model.
• To assess the utility of the proposed model for estimating MTR in MS lesions.

Main Methods:

• Lag functional linear model incorporating multiple functional predictors.
• Generalized cross-validation for time-specific smoothness.
• Restricted maximum likelihood for global smoothing.
• Numerical studies to evaluate predictive accuracy.
• Application to estimate MTR in multiple sclerosis lesions using MRI data.

Main Results:

• The proposed lag functional linear model effectively estimates MTR in MS lesions.
• Both generalized cross-validation and restricted maximum likelihood approaches yield accurate estimations.
• The model demonstrates superior performance compared to cross-sectional models lacking temporal considerations.
• The method is valuable for retrospective research and potential clinical use.

Conclusions:

• The lag functional linear model offers a robust method for MTR estimation in MS lesions.
• This approach enhances the utility of standard imaging modalities for assessing tissue damage.
• The model's ability to account for temporal dynamics improves predictive accuracy in neurological disease research.