Skip to content

fmri_data_dual_regression.md

Latest commit

 

History

History
82 lines (61 loc) · 3.29 KB

fmri_data_dual_regression.md

File metadata and controls

82 lines (61 loc) · 3.29 KB
Raw

fmri_data.dual_regression — FSL-style dual regression of group maps onto a subject's data

← back to fmri_data methods · Object methods index · Recasting objects

Project a set of group-level spatial maps (e.g., group ICA components) onto a single subject's 4-D data to recover (a) per-component time courses and (b) subject-specific spatial maps. The first regression is spatial (maps against each volume) and the second is temporal (recovered time courses against each voxel) — the standard FSL dual-regression workflow.

Code map

dual_regression code map

Editable PowerPoint version

Usage

[spatial_maps, timecourses, tmaps] = dual_regression(group_map_obj, data_obj, varargin)

The method is dispatched on group_map_obj: the receiver carries the group-level component maps and data_obj is the subject's 4-D image series. data_obj is resampled into the space of group_map_obj before the regressions are performed; the heavy lifting is delegated to dual_regression_fsl.

Inputs

Argument Type Description
group_map_obj fmri_data Group-level spatial maps, one image per component (e.g., from group ICA).
data_obj fmri_data Subject's 4-D BOLD time series. Resampled to the space of group_map_obj.
varargin misc. Forwarded to dual_regression_fsl.

Outputs

Output Type Description
spatial_maps fmri_data Subject-specific spatial maps, one image per component.
timecourses matrix [components × timepoints] recovered time courses.
tmaps fmri_data Z-scored version of spatial_maps (effect-size maps).

Notes

  • Resampling happens with resample_space(data_obj, group_map_obj) — the subject data are aligned to the group-map grid, not the other way around.
  • All real work is performed by dual_regression_fsl; this method exists to provide the standard CANlab object-method dispatch and to handle space alignment automatically.
  • Pair this with load_image_set('npsplus'), ICA component sets, or any other reference map collection of interest.

Example: project a small ICA component set onto a subject's run

% Use a few CANlab signature maps as stand-in components
group_map_obj = load_image_set('npsplus');
group_map_obj = get_wh_image(group_map_obj, 2:7);
disp(group_map_obj.image_names)

% Single-subject 4-D BOLD data
data_obj = fmri_data(which('swrsub-sid001567_task-pinel_acq-s1p2_run-03_bold.nii.gz'));

% Dual regression
[spatial_maps, timecourses, tmaps] = dual_regression(group_map_obj, data_obj);

% Inspect the recovered time courses
plot(timecourses');
xlabel('Volume'); ylabel('Component score');

See also