by Davor Curic
Physics PhD student – Supervisor: Joern Davidsen
University of Calgary, Alberta, Canada
Email me!
The critical brain hypothesis posits that in an effort to maximize information transfer, the brain self-organizes to a critical point which is sensitive to perturbations, but does not allow for never-ending growth of activity. This framework has largely been motivated by scale-free statistics observed in on-going spontaneous neuronal activity patterns across many scales, from cellular to fMRI, and in many species. Despite this, the presence of power-law distributions in the context of non-spontaneous activity (i.e., either externally or self evoked) remains controversial. Some studies suggest that evoked dynamics exhibit non-scale-free avalanches, while others suggest that the non-stationary levels of activity necessitate the use of adaptive binning/ thresholding methods in order to observe scale-free statistics in evoked activity. However, many of these studies do not take into account the variable susceptibility that different cells have to particular inputs. In this study, we investigated a region of the retrosplenial cortex of head-fixed mice running on a track using two-photon calcium imaging with single cell resolution. Within this region we identified three distinct groups; cells which are suppressed during periods of running, cells which increases activity during running periods, and a third group which is invariant to the behavioural state of the animal. We study the effect of mixing the distinct dynamics of these groups and how their individual avalanche statistics contribute to the whole population level. We also study how the effect of a psychedelic drug (ibogaine) changes the total population dynamics and how this relates to the dynamics of the distinct subgroups.
The critical brain hypothesis posits that in an effort to maximize information transfer, the brain self-organizes to a critical point which is sensitive to perturbations, but does not allow for never-ending growth of activity.
Despite the ubiquity of power-law statistics in spontaneous neural activity (a prerequisite for the critical state), the presence of power-law distributions in the context of non-spontaneous activity (i.e., either externally or self evoked) remains controversial.
Some studies suggest that evoked dynamics exhibit non-scale-free avalanches, while others suggest that the non-stationary levels of activity necessitate the use of adaptive binning/ thresholding methods in order to observe scale-free statistics in evoked activity.
Most experiments studying avalanches in task-based settings do not take into account the variable susceptibility that different cells have to particular inputs.
In this study, we investigated a region of the retrosplenial cortex of head-fixed mice running on a track using two-photon calcium imaging with single cell resolution (§ Experiment).
Within this region we identified three distinct groups; cells which are suppressed during periods of running (group S), cells which increases activity during running periods (group R), and a third group which is invariant to the behavioural state of the animal (group I). The results shown in the poster correspond to roughly 15% of the total population per group (§ Variable Activity Coding of Cells).
The distinct dynamics of these different cell groups are mixed together when the total population signal is studied. By defining activity-conditioned neuronal avalanches (§ Defining Neuronal Avalanches), we see that the different subgroups can have completely different activity-conditioned avalanche statistics, as indicated by the calculations of the avalanche statistic slopes. Specifically, while the R group resembles the total population statistics, the S and I groups are markedly different (§ Results).
Understanding exactly how different sub-populations contribute to the total population avalanche statistics is crucial in understanding neural avalanching behaviour in general.
This page only contains supplementary figures and clarifications pertinent to figures on the first page.
Hi, really liked your work on using 2PI to study psychedelics such as ibogaine in retrosplenial cortex (which is our lamp post for dorsal 2PI ;-)). Dividing your population by an avalanche independent measure is an important direction as we shouldn’t assume that every spike happens in the context of an avalanche. One would expect a brain to devote additional resources to control avalanche initiation, propagation etc.
Just want to point out some pitfalls in the history of dividing neuronal populations. E.g. in visual cortex, it was a tradition to focus on tuned neurons and disregard non-tuned. It is now very clear e.g. recent work by Jason MacLean Chicago that untuned neurons carry a lot of information, show important graphical motivs etc and that that division is largely arbitrarily set by the experimentalist. Similarly, in LFP reserach when you focus on a narrow bandwidth eg gamma only, you will miss the scaleinvariance visible in the broadband signal.
Best
Dietmar