Quality Control: DVARS and FD-DVARS correlations

The tmfc_calculate_DVARS function calculates DVARS within the GM mask before and after noise regression. It is called automatically by the main function TMFC_denoise if the user has selected the corresponding option, or it can be run manually:

[preDVARS,postDVARS] = tmfc_calculate_DVARS(FD,SPM_paths,options,masks,output_paths);

The outputs are saved in TMFC_denoise/[WM*e*]_[CSF*e*]_[GM*d*] subfolders:

DVARS_before_denoising.mat – contains DVARS time series for each session before noise regression, session-wise FD-DVARS and task-DVARS correlations, and mean/max FD-DVARS (and task-DVARS) correlation across sessions.

DVARS_*.mat – contains the same information as the previous file, but after noise regression. Filenames encode the selected denoising options (e.g., DVARS_[24HMP]_[aCompCor50]_[rWLS].mat).

The GUI window for DVARS time-series inspection is opened with tmfc_plot_DVARS. It is called automatically by the main function TMFC_denoise.

_images/DVARS_plot.svg — Graphical interface for DVARS time-series inspection. Example DVARS plot for a single subject. The FD-DVARS correlation was reduced from 0.69 to 0.03 after noise regression. Spikes in the DVARS time series during the first session, associated with high-motion events, were visibly diminished. At the group level, the mean FD-DVARS correlation was decreased toward zero.

To open DVARS plot GUI manually run:

% Allows saving group FD-DVARS statistics only:
tmfc_plot_DVARS(preDVARS,postDVARS,FD);

% Allows saving group FD-DVARS statistics and TMFC denoise settings:
tmfc_plot_DVARS(preDVARS,postDVARS,FD,options,SPM_paths,subject_paths,anat_paths,func_paths,masks);

Pressing the Save button stores individual subject FD and DVARS data, as well as group-wise DVARS statistics, in a single *.mat file:

Group_FD_DVARS.mat file
Field	Description
denoising_settings (struct)	Selected TMFC_denoise settings: SPM_paths: see Select Subjects subject_paths: see Select Subjects options: see Denoising Options anat_paths: see Select Structural Images func_paths: see Select Functional Images masks: see Mask Generation
FD (struct)	Individual FD data for all subjects (see HMP expansions and FD plots).
group_mean_post_FD_DVARS_corr	Group mean FD-DVARS correlation after denoising.
group_mean_pre_FD_DVARS_corr	Group mean FD-DVARS correlation before denoising.
group_SD_post_FD_DVARS_corr	Group SD of FD-DVARS correlation after denoising.
group_SD_pre_FD_DVARS_corr	Group SD of FD-DVARS correlation before denoising.
postDVARS (struct)	DVARS data for each subject (after denoising). Sess: Incluse DVARS time series, FD-DVARS and task-DVARS correlations for each session. Mean_FD_DVARS_corr: Mean FD-DVARS correlation across sessions. Max_FD_DVARS_corr: Maximum FD-DVARS correlation across sessions. taskDVARS_corr_mean: Mean task-DVARS correlation across sessions. taskDVARS_corr_maxabs: Maximum abs(task-DVARS correlation) across sessions. taskDVARS_corr_maxabs_name: Name of the corresponding task condition.
preDVARS (struct)	DVARS data for each subject (before denoising). Sess: Incluse DVARS time series, FD-DVARS and task-DVARS correlations for each session. Mean_FD_DVARS_corr: Mean FD-DVARS correlation across sessions. Max_FD_DVARS_corr: Maximum FD-DVARS correlation across sessions. taskDVARS_corr_mean: Mean task-DVARS correlation across sessions. taskDVARS_corr_maxabs: Maximum abs(task-DVARS correlation) across sessions. taskDVARS_corr_maxabs_name: Name of the corresponding task condition.

These values can be reported to demonstrate the effectiveness of noise regression. If denoising is successful, spikes in the DVARS time series at high-motion time points should be reduced, and the FD-DVARS correlation should approach zero.