Dear Carolyn,
On Mon, 29 Jul 2019 14:22:47 +0100, Carolyn <[log in to unmask]> wrote:
>Dear CAT12 experts,
>
>Because "Analyze ROIs" examines ALL ROIs of an atlas across the entire brain, does that mean the FDR-corrected results adjusts for total number of ROIs in the atlas?
The FDR and Bonferroni-Holmes correction is applied by including all regions. The easiest way to correct for fewer regions is to use the uncorrected p-value and adjust manually with Bonferroni or Bonferroni-Holmes. The latter is correcting the lowest p-value by n (number of regions) and the second lowest by n-1 and so on and is less conservative.
>
>To restrict to a small number of a priori ROIs, do we use the "Estimate mean values inside ROI" option instead?
>(Taking the values from only the e.g. 3xROIs columns that one wants to select, to do further analysis in external programmes (R/JASP/SPSS etc.) and adjust for multiple tests with regard to the only the 3 ROIs)?
This is also an alternative. However, you have to keep in mind that no nuisance paramaters (e.g. TIV) are considered and these are the ROI values from native data.
>
>I have seen papers that after whole brain voxel-wise analysis, have "calculated masks according to Neuromorphometric Atlas and performed additional small volume correction.." (e.g. Besteher et al. 2019, 2017).
>I assume this means creating mask images of selected ROIs from the atlas (e.g. using spm_atlas and spm_write_vol functions), and then in spm results, selecting "small volume correction" with the ROI mask image as Search Volume?
Yes, this will work.
>In this approach, is this still detecting voxel-wise group differences, rather than averages across all voxels of an atlas-based ROI to detect differences regardless of whether the voxels overlap or not, which is what the estimated mean values csv files contain? (e.g. Nickel et al. 2017) Or are these two approaches similar?
No, the SVC is a voxel-wise approach, but restricted to some predefined regions. The ROI approach above is a ROI-wise approach.
>
>Could one also use the inclusive mask function of SPM results, selecting Atlas>Neuromorphometrics, select the ROI from the list labels?
Again, this will be a voxel-wise approach.
Best,
Christian
>
>Is one of these methods the most appropriate?
>
>Thank you for your time.
>
>Best regards,
>Carolyn
>
>
>
>
>Besteher, B., Gaser, C., & Nenadić, I. (2019). Brain structure and trait impulsivity: A comparative VBM study contrasting neural correlates of traditional and alternative concepts in healthy subjects. Neuropsychologia, 131, 139–147. https://doi.org/10.1016/j.neuropsychologia.2019.04.021
>
>Besteher, B., Squarcina, L., Spalthoff, R., Bellani, M., Gaser, C., Brambilla, P., & Nenadić, I. (2017). Brain structural correlates of irritability: Findings in a large healthy cohort: Brain Structural Correlates of Irritability. Human Brain Mapping, 38(12), 6230–6238. https://doi.org/10.1002/hbm.23824
>
>Nickel, K., Joos, A., Tebartz van Elst, L., Matthis, J., Holovics, L., Endres, D., … Maier, S. (2018). Recovery of cortical volume and thickness after remission from acute anorexia nervosa. International Journal of Eating Disorders, 51(9), 1056–1069. https://doi.org/10.1002/eat.22918
|