A fast and general method to empirically estimate the complexity of distributed causal interactions in the brain
BackgroundThe Perturbational Complexity Index (PCI) was recently introduced to assess the capacity of thalamocortical circuits to engage in complex patterns of causal interactions. While showing high accuracy in detecting consciousness in brain injured patients, PCI depends on elaborate experimental setups and offline processing and has restricted applicability to other types of brain signals beyond transcranial magnetic stimulation and high-density EEG (TMS/hd-EEG) recordings. ObjectiveWe aim to address these limitations by introducing PCIST, a fast method for estimating perturbational complexity of any given brain response signal. MethodsPCIST is based on dimensionality reduction and state transitions (ST) quantification of evoked potentials. The index was validated on a large dataset of TMS/hd-EEG recordings obtained from 108 healthy subjects and 108 brain injured patients, and tested on sparse intracranial recordings (SEEG) of 9 patients undergoing intra-cerebral single-pulse electrical stimulation (SPES). ResultsWhen calculated on TMS/hd-EEG potentials, PCIST performed with the same accuracy as the original PCI, while improving on the previous method by being computed in less than a second and requiring a simpler set-up. In SPES/SEEG signals, the index was able to quantify a systematic reduction of intracerebral complexity during sleep, confirming the occurrence of state-dependent changes in the effective connectivity of thalamocortical circuits, as originally assessed through TMS/hd-EEG. ConclusionsPCIST represents a fundamental advancement towards the implementation of a reliable and fast clinical tool for the bedside assessment of consciousness as well as a general measure to explore the neuronal mechanisms of loss/recovery of brain complexity across scales and models.