Dear Cameron,

thanks very much for making the atlas and scripts available. I have some questions regarding this. I'd like to use the parcellations as independent ROIs for our task-based fMRI study to extract parameter estimates and correlate these with behavioural measures. I understand that a cluster size of 200 does still represent meaningful anatomical labels. I'll also try higher splitting, but in general would that be a sound approach (using these as independent ROIs)?

From reading your article I don't think you compared the result s you've obtained from your sample with the independent one (Biswal) in terms of consistency across different study populations when running the tcorr. Or did you and it's just not been reported?

Since the parcellation is based on timeseries similarity, there's no way of attributing a cluster to a certain functional network as size and thus overlap will depend on the number of clusters chosen, is that right?

Thanks very much for your help and efforts.

Best,

Torsten

Hello Thorsten,

Sorry for the late reply.

I think that the analysis that you propose is valid.

I did not look at consistency across different populations, but this is a very interesting thing to do.

The size of an ROI and its overlap with regions of other atlases will depend on size, but you can still attribute clusters to functional networks. For example you could assign regions to networks derived in Yeo and Buckner 2011 by amount of overlap.

I hope these help!

Cheers,
Cameron

Hi Cameron,

thanks very much for your response - I'll look into that Yeo et al. paper.

Cheers,

Torsten