Inducing a meditative state by artificial perturbations: A mechanistic understanding of brain dynamics underlying meditation

Abstract Contemplative neuroscience has increasingly explored meditation using neuroimaging. However, the brain mechanisms underlying meditation remain elusive. Here, we implemented a mechanistic framework to explore the spatiotemporal dynamics of expert meditators during meditation and rest, and controls during rest. We first applied a model-free approach by defining a probabilistic metastable substate (PMS) space for each condition, consisting of different probabilities of occurrence from a repertoire of dynamic patterns. Moreover, we implemented a model-based approach by adjusting the PMS of each condition to a whole-brain model, which enabled us to explore in silico perturbations to transition from resting-state to meditation and vice versa. Consequently, we assessed the sensitivity of different brain areas regarding their perturbability and their mechanistic local-global effects. Overall, our work reveals distinct whole-brain dynamics in meditation compared to rest, and how transitions can be induced with localized artificial perturbations. It motivates future work regarding meditation as a practice in health and as a potential therapy for brain disorders.

cluster centroids Vc(t) rendered onto brain maps.The leading communities found for each substate similarly overlapped the ones revealed for k=5 of the main analysis, highlighting the robustness of the empirical LEiDA approach for different numbers of cluster centers.Furthermore, it reveals the clear characterization of resting-state and meditation brain dynamics in expert meditators.Metastable substate 1 was characterized by all negative eigenvector elements, as in substate 1 of k=5.In addition, substate 2 was led mainly by areas from the DMN, and very few from the control network and limbic system.Furthermore, substate 3 had a community dominated by the somatomotor and salience networks.Here, substate 2 and substate 3 could be related with substate 3 and substate 2 from k=5, respectively.This relation differs in terms of the overall probability of occurrence in each PMS: the substate led by the DMN had a higher probability of occurrence in k=4, whereas the substate led by the somatomotor and salience network has a higher probability of occurrence in k=5).The last substate 4 had a functional network led by the visual system and the somatomotor network, and could be closely related with substate 5 in k=5.The probability of occurrence of meditation compared to resting-state was significantly lower in substate 3 [0.1410± 0.0132 vs. 0.1924 ± 0.0177, P=0.0063].Model-based results: Fitting and optimization for each brain state.In the analysis of expert meditators during resting-state and meditation for k=5, a in rest G=0.04 (KLD=0.0064)and b in meditation G=0.05 (KLD=0.0122).In the analysis of expert meditators during resiting-state and meditation for k=4, c in rest G=0.04 (KLD=0.0058)and d in meditation G=0.04 (0.0141).In the analysis of controls during resiting-state and expert meditators during meditation for k=2, e in rest G=0.04 (KLD=0.0003)and f in meditation G=0.04 (0.0002).

Figure. S 1 .
Figure.S 1. LEiDA for k=4 in the analysis of expert meditators during resting-state and meditation.a Empirical Probabilistic Metastable Substate

Figure. S 3 .
Figure.S 3. Correlation between the probability of occurrence of each substate from the PMS and experience of expert meditators.No significant correlation was found.a LEiDA on expert meditators during resting-state and meditation for k=5, PMS of rest.b LEiDA on expert meditators during restingstate and meditation for k=5, PMS of meditation.c LEiDA on expert meditators during resting-state and meditation for k=4, PMS of rest.d LEiDA on expert meditators during resting-state and meditation for k=4, PMS of meditation.e LEiDA on controls during resting-state and expert meditators during meditation for k=2, PMS of meditation.

Figure. S 4 .
Figure.S 4. Correlation between scanning motion and substate occupancy.Correlation between the condition delta framewise displacement (meditation

Figure. S 5 .
Figure.S 5. Static analysis of expert meditators during resting-state and meditation.Functional connectivity is not sensitive enough to characterize differences between the two brain states.a Histogram of the mean FC matrix of rest and mean FC matrix of meditation.The distributions closely overlap.b Pvalues of the node-to-node comparison of the FC matrices of all subjects in each brain state.None survive correction by multiple comparisons.c Visualization

Figure. S 7 .
Figure.S 7. Model-based results for k=4: Whole-brain models in the analysis of expert meditators during resting-state and meditation.Empirical and simulated PMS for a resting state (G=0.04 with a KL distance of 0.0058) and b meditation (G=0.04 with a KL distance of 0.0141).

Figure. S 9 .
Figure.S 9. Model-based results for k=4: In silico stimulation in the analysis of expert meditators during resting-state and meditation.This shows the robustness of the perturbational approach for different values of k in regards to the identification of the most sensitive brain areas to promote a transition from resting-state to meditation in expert meditators and vice versa.a Transitions from rest towards meditation were possible for the synchronization protocol, characterized by lower KL distance between the perturbed modeled PMS of rest and the empirical PMS of meditation.b In the synchronization protocol, optimal perturbation for each brain area.The color bar represents the KL distances.c The best transition was found in the area LH Som 3 from Schaeffer 100 parcellation (Schaefer et al., 2018) for a bifurcation parameter value of a=0.11.The perturbed rest PMS is closer to the target empirical meditation PMS.d Opposite transitions from meditation to rest were found in the noise protocol.e Optimal perturbation of each brain area in the noise protocol, at their particular simulation intensities, rendered onto brain maps.f Best transition was obtained at a bifurcation parameter value of a=-0.14 in the area LH Default PCunCC 2