A longitudinal model for functional connectivity networks using resting-state fMRI.

Area of Science:

• Neuroimaging
• Neuroscience
• Biostatistics

Background:

• Longitudinal functional magnetic resonance imaging (fMRI) studies are common, but lack a unified framework for analyzing functional connectivity (FC) networks over time.
• Existing methods struggle to account for within-subject dependence, population variability, and autocorrelation in fMRI time series data.

Purpose of the Study:

• To develop a novel longitudinal FC model using a variance components approach for analyzing longitudinal fMRI data.
• To provide a computationally feasible and stable method that accounts for within-subject dependence, population variability, and autocorrelation.
• To enable valid inference on group differences in baseline FC and longitudinal changes in FC.

Main Methods:

• Employed a variance components approach to model longitudinal FC networks.
• Utilized a non-parametric technique to address autocorrelation in fMRI time series.
• Applied generalized least squares to estimate within-subject variance, baseline FC, and longitudinal FC trends.
• Developed a permutation testing procedure for statistical inference on group differences.

Main Results:

• The novel longitudinal FC model was applied to data from the Alzheimer's Disease Neuroimaging Initiative (ADNI).
• No significant differences were found in global FC networks between Alzheimer's disease patients and healthy controls.
• Significant differences were observed in local aging patterns of FC between the left hippocampus and posterior cingulate cortex.

Conclusions:

• The proposed variance components model offers a robust framework for analyzing longitudinal FC in fMRI data.
• The findings suggest that while global FC may be preserved, specific local connectivity patterns exhibit altered aging in Alzheimer's disease.
• This approach advances the analysis of neuroimaging data in longitudinal studies, particularly for neurodegenerative diseases.