Linking Covariates to State Durations in Dynamic Brain Networks via Hidden Semi-Markov Models

Covariate-dependent Hidden semi-Markov models link age and sex to dwell times in a subset of brain states in resting-state fMRI from 1,000 UK Biobank participants.

• Why it matters: Dwell times are often only compared indirectly after fitting, which restricts formal covariate modeling and makes it hard to include multiple predictors simultaneously.
• What they did: The authors add a regression formulation to HSMM dwell-time distributions, estimate parameters with expectation-maximization, and quantify dwell-time regression uncertainty via a subject-level nonparametric bootstrap.
• The result: In simulations, the method accurately recovered covariate effects under shifted-Poisson and Gamma dwell-time specifications, and in UK Biobank data it shows age and sex are associated with state persistence.

AI summary of the abstract · may contain errors · not medical advice · read the paper.