Quantifying the unpredictable: Entropy as a window into resting-state brain dynamics

Resting-state functional magnetic resonance imaging (rs-fMRI) is widely used to investigate intrinsic brain organisation and spontaneous neural dynamics. Entropy-based methods have emerged as important tools for quantifying signal complexity, uncertainty, and information processing in rs-fMRI data. This systematic review examines the application of five entropy measures, including approximate entropy, sample entropy, multiscale entropy, Shannon entropy, and transfer entropy, in resting-state fMRI research. A systematic search of the Web of Science database identified studies published between 2000 and 2025, of which 111 met inclusion criteria for methodological analysis. Studies were categorised by entropy measure, analytical scope, and subject population. Applications clustered by analytical scope. Shannon entropy dominated connectivity-level analyses, while sample and multiscale entropy were favoured for regional complexity. Transfer entropy was applied primarily in studies of directed connectivity. Across studies, entropy-based metrics demonstrated sensitivity to age and pharmacological modulation, with entropy alterations frequently mapping onto established functional brain networks. However, substantial heterogeneity was observed in preprocessing strategies, parameter selection, and statistical inference, limiting cross-study comparability. The evidence supports entropy as a flexible framework for characterising resting-state brain dynamics while highlighting significant methodological challenges related to standardisation and reproducibility. Future research should prioritise harmonisation of entropy parameterisation, validation against null models, and broader application in clinical populations.