Skip to content

Latest commit

 

History

History
111 lines (86 loc) · 5.55 KB

output.rst

File metadata and controls

111 lines (86 loc) · 5.55 KB
Raw

Output

There are two classes of outputs returned by fMRIDenoise:

1. Quality measures report: fMRIDenoise returns a group-level .html report, where you can find details about the performance of each denoising strategy.

2. Denoised fMRI data: fMRIDenoise denoises your preprocessed fMRI data with selected denoising strategies and stores individually denoised data within <output dir>/derivatives/fmridenoise/sub-<label> folders. You can use this data for your further analysis.

Quality measures report

fMRIDenoise returns a summary report, written to <output dir>/fmridenoise/fmridenoise_report.html (see an example fMRIDenoise report). The report contains various quality measures that might help to decide which denoising strategy perform best on the particular data. Here we list all quality measures that are calculated when running the pipeline. The raw quality measures calculated for each strategy are also stored in .tsv tables in the main <output dir>/derivatives/fmridenoise folder.

FC-FD Pearson correlation

Effective denoising of fMRI signals should result in minimization of a statistical relationship between head motion and functional connectivity estimates [Parkes2018]. FC-FD Pearson correlation metric represents the relationship between individual head motion and functional connectivity, calculated as the Pearson correlation between subject-specific framewise displacement (FD) [Power2012] and functional connectivity calculated for each edge. fMRIDenoise reports both median absolute FC-FD Pearson correlation as well as the proportion of edges for which this correlation was statistically significant (p < 0.05, uncorrected) [Parkes2018]. FC-FD Pearson correlation metrics are reported both for all subjects and for the subgroup of subjects with a low head motion.

FC-FD distance-dependence

Head motion in the scanner might inflate the strength of short-distance connections when compared to medium- and long-distance connections [Power2012]. Effective denoising of fMRI signals should result in no distance-dependence of FC-FD Pearson correlations. FC-FD Distance dependence metric represents the Spearman correlation between FC-FD correlations and the euclidean distance calculated between brain regions [Parkes2018]. Distance dependence is reported both for all subjects and for the subsample of subjects with low head motion.

tDOF-loss

Denoising pipelines may vary in the number of regressors used to model the noise in fMRI timeseries. Including more nuisance regressors in the denoising procedure could result in improving the modeling a non-neuronal signal noise, but also and in a loss of temporal degrees of freedom (tDOF-loss) [Parkes2018]. Reduced degrees of freedom could result in spuriously increased connectivity estimates. Additionally, applying volume censoring and ICA-AROMA could drive artificial variance in functional connectivity estimates [Prium2015]. Loss of temporal degrees of freedom (tDOF-loss) metrics represents the number of regressors included in the denoising procedure.

Edge density

Connectivity matrices calculated based on noisy signals characterize in highly positively-skew distribution of edges' weights. Report includes the edges weight distributions for networks obtained for each denosing strategy.

Excluded subjects

The list of subject that should be excluded from further data analysis due to high motion (mean FD > 0.2 or max FD > 5 or more than 10% od outlier data points).

Denoised fMRI data

Denoised data are written to <output dir>/fmridenoise/sub-<subject_label>/, following BIDS naming convention:

sub-<subject_label>/
       ├── sub-<subject_label>_task-<task_label>_pipeline-24HMP8PhysSpikeReg_carpetPlot.png
       ├── sub-<subject_label>_task-<task_label>__pipeline-24HMP8PhysSpikeReg_connMat.npy
       ├── sub-<subject_label>_task-<task_label>_pipeline-24HMP8PhysSpikeReg_desc-confounds.tsv
       ├── sub-<subject_label>_task-<task_label>_space-MNI2009cAsym_pipeline-24HMP8PhysSpikeReg_desc-denoised_bold.nii.gz

Content:

  • carpetPlot.png - carpet plot representing timeseries before and after denoising
  • connMat.npy - correlation matrix calculated based on denoised data
  • confounds.tsv - filtered confounds table used for selected denoising pipeline
  • denoised_bold.nii.gz - denoised fMRI data

The same files structure is generated for each denoising pipeline.

References