[DCAN] QCPlot shows FD after filtering

[tsalo]

Summary

Based on the code below, it looks like QCPlot uses the filtered framewise displacement values in both the preprocessing and postprocessing carpet+line plots. This is probably misleading to users, since it seems like no filtering was done between the two versions.

xcp_d/xcp_d/interfaces/qc_plot.py

Lines 104 to 119 in d4f8bd0

	confound_matrix = load_confound(datafile=self.inputs.bold_file)[0]
	motion_conf = load_motion(
	confound_matrix.copy(),
	TR=self.inputs.TR,
	motion_filter_type=self.inputs.motion_filter_type,
	motion_filter_order=self.inputs.motion_filter_order,
	freqband=[self.inputs.low_freq, self.inputs.high_freq])
	# Pull out motion confounds
	motion_df = pd.DataFrame(data=motion_conf.values,
	columns=[
	"rot_x", "rot_y", "rot_z", "trans_x",
	"trans_y", "trans_z"
	])
	# Compute fd_timeseries from motion_confounds df
	fd_timeseries = compute_fd(confound=motion_df,
	head_radius=self.inputs.head_radius)

[tsalo]

It might also be a good idea to show the power spectrum of the FD trace before and after filtering.

[tsalo]

@madisoth corrected me on this- we don't want to show unfiltered FD for preprocessed data and filtered for postprocessed data, because the filtered FD is only used for censoring; it's not used for denoising. As such, we want to show the unfiltered FD in both plots. #579 adds a new censoring plot to the summary reports, so users should be less confused.

[tsalo]

As it turns out, they DCAN team does want us to use the filtered FD for denoising. Not sure why they wanted us to only show unfiltered FD in the plots.