help
help > RE: Denoising and V2V doubts
Dec 10, 2019 01:12 PM | Alfonso Nieto-Castanon - Boston University
RE: Denoising and V2V doubts
Dear Inés,
Regarding (1), yes, both linear detrending and high-pass filtering are aimed at removing very-low-frequency effects, yet they are not redundant, high-pass filtering alone would not fully remove linear trends in the data and linear detrending would not remove other non-purely-linear low-frequency fluctuations, so applying both (as CONN does by default, together with removing of other motion, scrubbing, and White/CSF aCompCor factors) helps more thoroughly denoise your data.
Regarding (2), yes, CONN implements Martuzzi's ICC-power measure (also referred to as ICC-p0 in the original Martuzzi's 2011 manuscript), which involves no thresholding operation.
Hope this helps
Alfonso
Originally posted by Ines Del Cerro:
Regarding (1), yes, both linear detrending and high-pass filtering are aimed at removing very-low-frequency effects, yet they are not redundant, high-pass filtering alone would not fully remove linear trends in the data and linear detrending would not remove other non-purely-linear low-frequency fluctuations, so applying both (as CONN does by default, together with removing of other motion, scrubbing, and White/CSF aCompCor factors) helps more thoroughly denoise your data.
Regarding (2), yes, CONN implements Martuzzi's ICC-power measure (also referred to as ICC-p0 in the original Martuzzi's 2011 manuscript), which involves no thresholding operation.
Hope this helps
Alfonso
Originally posted by Ines Del Cerro:
Dear CONN users,
Recently we received the comments of a reviewer and I am having some problems to find some information about CONN processing:
1. Linear detrending and band-pass filtering: we receive the comment that they both remove low-frequencies, but even if it is true, the regression that linear detrend does is not performed by applying a pass filter. Is that the reason why both are included in the CONN's default denoising pipeline?
2. ICC analysis: The reviewer ask us for the correlation threshold of ICC. In Martuzzi et al., 2011 there are described two different methods (ICC-degree and ICC-power). Given that the ICC-power does not required for a previous threshold, I am guessing that is the one that CONN applies. Am I right?
Thanks in advance,
Inés
Recently we received the comments of a reviewer and I am having some problems to find some information about CONN processing:
1. Linear detrending and band-pass filtering: we receive the comment that they both remove low-frequencies, but even if it is true, the regression that linear detrend does is not performed by applying a pass filter. Is that the reason why both are included in the CONN's default denoising pipeline?
2. ICC analysis: The reviewer ask us for the correlation threshold of ICC. In Martuzzi et al., 2011 there are described two different methods (ICC-degree and ICC-power). Given that the ICC-power does not required for a previous threshold, I am guessing that is the one that CONN applies. Am I right?
Thanks in advance,
Inés
Threaded View
| Title | Author | Date |
|---|---|---|
| Ines Del Cerro | Dec 10, 2019 | |
| Alfonso Nieto-Castanon | Dec 10, 2019 | |