Characterizing patterns of diffusion tensor imaging variance in aging brains.

Area of Science:

• Neuroimaging
• Biostatistics
• Medical Physics

Background:

• Large-scale neuroimaging studies require robust statistical methods to handle data variance.
• Diffusion Tensor Imaging (DTI) is susceptible to spatially varying noise, necessitating careful consideration of distributional assumptions.
• Understanding sources of variability in DTI metrics is critical for accurate interpretation of results, especially when merging data from multiple sites.

Purpose of the Study:

• To characterize the role of physiological factors, subject compliance, and scanner interactions in DTI variability.
• To model DTI variability by analyzing the spatial variance of derived metrics within homogeneous regions.
• To investigate how covariates such as age, session interval, and motion influence DTI variance across different brain regions.

Main Methods:

• Analysis of DTI data from 1035 subjects in the Baltimore Longitudinal Study of Aging (BLSA).
• Assessment of DTI scalar variance within regions of interest (ROIs) defined by four segmentation methods.
• Investigation of relationships between DTI variance and covariates including age, time from baseline, motion, sex, and scan session number.

Main Results:

• Covariate effects on DTI variance are heterogeneous and bilaterally symmetric across ROIs.
• Inter-session interval, sex, and head motion significantly influence fractional anisotropy (FA) variance in specific brain regions.
• Head motion increases during rescans, indicating potential changes in subject compliance or scanner interaction over time.

Conclusions:

• The influence of covariates on DTI variance is complex and region-specific.
• Researchers are encouraged to report variance estimates and consider models of heteroscedasticity in DTI analyses.
• This study provides a foundation for planning future studies to account for regional variations in DTI metric variance.