help
help > RE: high-pass filter
Feb 4, 2021 08:02 PM | Alfonso Nieto-Castanon - Boston University
RE: high-pass filter
Dear Bram,
That is strange (and yes, the confound-corrected timeseries does include any band-pass filtering specified in the Denoising step).In your plots (assuming the y-axis is in dB units) the before-denoising timeseries already looks rather unusual (the power looks strangely constant below 0.010Hz and then drops dramatically above 0.010Hz). I am thinking perhaps this could be caused by concatenating the timeseries from different sessions or some other similar artifact when computing the power spectrum... Could you perhaps share a few more details about your data, any previous preprocessing, and on how you compute these global BOLD signals and their power spectra?
Best
Alfonso
Originally posted by Bram Beckers:
That is strange (and yes, the confound-corrected timeseries does include any band-pass filtering specified in the Denoising step).In your plots (assuming the y-axis is in dB units) the before-denoising timeseries already looks rather unusual (the power looks strangely constant below 0.010Hz and then drops dramatically above 0.010Hz). I am thinking perhaps this could be caused by concatenating the timeseries from different sessions or some other similar artifact when computing the power spectrum... Could you perhaps share a few more details about your data, any previous preprocessing, and on how you compute these global BOLD signals and their power spectra?
Best
Alfonso
Originally posted by Bram Beckers:
Dear Alfonso,
I would like to apply a high-pass filter to my data, I have therefore set the band pass to [0.008 inf] (applying after regression, empty confound list, no despiking, no detrending).
In the options tab I have checked the 'create confound corrected timeseries' box.
Afterwards, I extracted the global signal from both the raw data and the corrected timeseries. When subsequently computing the power spectrum for both timeseries, I am still getting a large peak in the very low frequency range. For comparison I used MATLABs highpass function on the raw global signal and computed the power spectrum in a similar way (see attached figure). The latter looks quite different, with very low frequencies being suppressed more clearly.
Is band pass filtering incorporated in the confound corrected timeseries? Or is the filter only applied within CONN (analyses) itself?
Many thanks in advance.
Best regards,
Bram
I would like to apply a high-pass filter to my data, I have therefore set the band pass to [0.008 inf] (applying after regression, empty confound list, no despiking, no detrending).
In the options tab I have checked the 'create confound corrected timeseries' box.
Afterwards, I extracted the global signal from both the raw data and the corrected timeseries. When subsequently computing the power spectrum for both timeseries, I am still getting a large peak in the very low frequency range. For comparison I used MATLABs highpass function on the raw global signal and computed the power spectrum in a similar way (see attached figure). The latter looks quite different, with very low frequencies being suppressed more clearly.
Is band pass filtering incorporated in the confound corrected timeseries? Or is the filter only applied within CONN (analyses) itself?
Many thanks in advance.
Best regards,
Bram
Threaded View
| Title | Author | Date |
|---|---|---|
| Bram Beckers | Feb 3, 2021 | |
| Alfonso Nieto-Castanon | Feb 4, 2021 | |
| Bram Beckers | Feb 8, 2021 | |
| Alfonso Nieto-Castanon | Feb 14, 2021 | |