Reliability of energy landscape analysis of resting-state functional MRI data

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Data Science

Background:

• Energy landscape analysis is a data-driven technique for analyzing multidimensional time series, notably functional magnetic resonance imaging (fMRI) data.
• This method models data dynamics using an Ising model, representing them as a noisy ball moving on a derived energy landscape.
• It has demonstrated utility in characterizing fMRI data in both healthy and diseased states.

Purpose of the Study:

• To evaluate the test-retest reliability of energy landscape analysis applied to fMRI data.
• To develop and apply a permutation test to compare within-participant versus between-participant reliability.
• To assess the reliability of a variational Bayesian method for estimating participant-specific energy landscapes.

Main Methods:

• Fitting an Ising model to fMRI data to derive energy landscapes.
• Constructing a permutation test to compare within-participant (test-retest) and between-participant reliability.
• Utilizing a variational Bayesian method for individualized energy landscape estimation.

Main Results:

• Energy landscape analysis demonstrated significantly higher within-participant reliability compared to between-participant reliability across four common indices.
• The variational Bayesian method for estimating participant-specific energy landscapes showed comparable test-retest reliability to the conventional likelihood maximization method.
• The proposed permutation test effectively distinguished between within- and between-participant reliability.

Conclusions:

• Energy landscape analysis exhibits robust test-retest reliability, particularly at the individual participant level.
• The developed methodology supports statistically controlled, individual-level energy landscape analysis for fMRI data.
• This work validates energy landscape analysis as a reliable tool for characterizing brain dynamics in health and disease.