CONN v.13n ReleasedPosted By: Alfonso Nieto-Castanon - Jul 8, 2013
Tool/Resource: CONN : functional connectivity toolbox
The latest revision of the CONN functional connectivity toolbox (v.13n) includes some interesting additions we would like to highlight:
1) Network based statistics (NBS). ROI-to-ROI second-level analyses now include the ability to use NBS to control for false positives at the family-wise level, which is particularly useful when analyzing the entire functional connectome. NBS is a great method proposed by Zalesky et al. (Zalesky, A., Fornito, A., Bullmore, E.T. 2010. Network-based statistic: Identifying differences in brain networks. NeuroImage, 53(4):1197-1207) that controls for family-wise error (FWE) in mass-univariate testing of a full ROI-to-ROI connectivity matrix. It is a non-parametric approach based on permutation tests that looks at the extent of interconnected effects when analyzing the ROI-to-ROI connectivity matrix.
We have integrated and adapted their original implementation (which you can find, and we highly recommend, at http://www.nitrc.org/projects/nbs) to work with any mixed within- and between- subjects design that the CONN toolbox allows, using a general permutation of residuals approach (Still, A. W. and White, A. P., 1981. The approximate randomization test as an alternative to the F test in analysis of variance. Br. J. Math. Stat. Psychol., 34, 243–252) that requires minimal user interaction. In addition to network-level corrections, we have also added similar seed-level corrections to give users additional options for family-wise error control when analyzing the entire connectome or when looking at multiple seeds simultaneously. We will be updating the manuals soon to detail the new options but in the meantime you can find more details about these in the FAQ (http://www.alfnie.com/software/conn#TOC-...)
2) There has been some recent and interesting discussion regarding the influence of some methodological choices during nuisance regression of physiological/movement effects in fcMRI (Hallquist, M.N., Hwang, K., Luna, B., 2013. The nuisance of nuisance regression: Spectral misspecification in a common approach to resting-state fMRI preprocessing reintroduces noise and obscures functional connectivity. NeuroImage, 82:208-225). This manuscript highlights how a somewhat common preprocessing pipeline (referred as BPreg; band-pass before regression) that applies temporal band-pass filtering to the BOLD signal timeseries before removing nuisance effects can lead to inappropriate control of physiological/movement effects. The CONN toolbox uses instead by default a procedure referred as RegBP (regression before band-pass), which the manuscript finds to have similar performance to their preferred method (referred as Simult; simultaneous regression and band-pass). While the differences between these two methods (RegBP and Simult) seem to be relatively minor, with perhaps slightly better control offered by the 'Simult' approach at the cost of perhaps lower specificity, we have now included both methods as options in the Preprocessing tab, and encourage users to try these out. We expect that the 'Simult' approach should offer better control in those cases where the nuisance regressors (e.g. estimated subject motion parameters) and the associated BOLD signal effects (e.g. movement-related artifacts) show different spectral characteristics, e.g. when their relation is best characterized as a non-zero-order filtering operation; but it could be suboptimal -not the minimum variance estimator, leading to lower sensitivity yet still valid results- if they show similar spectral characteristics, e.g. when their relation is best characterized as a purely zero-order linear operation. We have left for now RegBP as the default settings in the toolbox just for consistency with previous versions of the toolbox, but we will keep a close tab on this issue and update the default settings if the community consensus seems to lean towards a different option.
Last there are as usual a few other smaller additions such as bug fixes (thanks to Vincent Beliveau for suggesting an effective solution to the svd-convergence issue that appeared under some Matlab versions!) and other minor modifications (we have finally added an option for the users to modify the default GUI font size -under Help->GUI settings-, which seemed to be a popular request; we have also further customized and extended the new 'Connectome view' display in the ROI-to-ROI results gui). We hope you find these useful as well, and as always please feel free to post your questions/issues/thoughts/suggestions in the forum, and thank you all for your support!.