|
Activity Number:
|
562
- Regression Methods for Neuroimaging Data
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 11, 2022 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics in Imaging
|
|
Abstract #323228
|
|
|
Title:
|
Mixed Spline Modeling to Estimate the Onset of MS-Specific Thalamic Atrophy Using Brain MRI Scans
|
|
Author(s):
|
Steven Y Cen* and Christina J. Azevedo and Daniel Pelletier and Saeed Moazami and Mulugeta Gebregziabher
|
|
Companies:
|
University of Southern California and University of Southern California and University of Southern California and University of Southern California and Medical University of South Carolina
|
|
Keywords:
|
mixed spline;
MRI;
multiple sclerosis;
brain;
aging;
missing dat
|
|
Abstract:
|
In multiple sclerosis (MS), the true disease duration can be difficult to estimate because the onset of disease specific tissue loss visible on structural brain magnetic resonance images precedes the clinical onset. We propose mixed spline models to fit the trajectory of thalamic atrophy to determine the age when the spline curve trajectory of a MS patient starts to deviate from a hypothetical normal aging trajectory for this patient. To investigate the heterogeneity of the trajectories, 26 covariate settings and 12 spline settings were considered. The study was motivated by a real-life dataset of 520 MS patients with a mean of 4±1.5 annual scans. A large cross-sectional dataset of 2138 normal aging participants was used to augment longitudinal data with 5 repeated scans through multivariate adaptive regression splines. Model fits were assessed using external validation with 50 randomly selected MS patients and internal validation (AIC/BIC). Repeated measurement and intra-class correlations were used to assess the agreement between the model predicted and the observed value. Mixed restricted basis spline or truncated power functions with TOEPLIZ as G-side matrix were the best fit.
|
Authors who are presenting talks have a * after their name.
