Content of review 1, reviewed on July 14, 2019

The article describes border reconfiguration of the connectome under various conditions (rest / task). I've seen a talk about this during the OHBM conference 2019 in Rome and decided to check the article, this is a really interesting and, I think, important result. I'm not an expert on connectivity parcellation method, but the article outlines nicely the issue of using hard boundaries.

My only major criticism is that there is a lack on results/data, as only similarity/consistency is reported (which is super important) but readers are left to infer what the parcel size data look like.
- by how much parcel size change? say in percent of the initial atlas parcels ; it would be nice to see in figure 1 histograms of that and/or have a summary table
- what would be interesting to see for the 'stability' analysis is not the effect on similarity (fig 6) but the actual amount of reconfiguration, not just from the Yale data but also from the MSC data to have a curve of the amount of reconfiguration (y) as a function of the number of (atlas) parcels (x) and then the similarity of these curves

A few other comments here
- data preprocessing mention temporal smoothing, which is equivalent to remove high frequencies; what's the sigma about - the number of volumes? what would be the equivalent low pass? and what about high pass filter
- testing the effect of task ; supplementary material was not available in the preprint?? assuming linearity (and modeling with a few basis functions) why not using residuals of the GLM to see the effect of the task?
- the discussion / additional considerations could do by mentioning the approach discussed with references, ie ReHo and ICA

Source

    © 2019 the Reviewer (CC BY 4.0).

References

    Salehi, M., Greene, A. S., Karbasi, A., Shen, X., Scheinost, D., Constable, R. 2018. There is no single functional atlas even for a single individual: Parcellation of the human brain is state dependent. BioRxiv.