pipeline.cfg /.json

A first-level model estimation options file lists the details of the General Linear Model estimation procedure.  

These files can be stored in [.cfg] or [.json] format. 

Model estimation option files may have a combination of the following fields:

#functional_label: label of functional data to enter in first-level analysis

only when used in combination with "functional_label" preprocessing steps: enter Secondary Dataset label identifying the desired functional dataset (default: Primary Dataset, i.e. fully preprocessed functional data)

#functional_smoothinglevel: level of smoothing of functional data to enter in first-level analysis

only when used in combination with preprocessing pipelines which implement multiple smoothing steps: enter 0/1/2 (0 unsmoothed data; 1 minimally-smoothed data; 2 fully-smoothed data) 

#model_basis     : temporal response function

hrf / hrf+deriv / hrf+derivs / none : response function  (default: hrf+deriv)

hrf for hemodynamic response function only

hrf+deriv to add temporal derivative

hrf+derivs to add temporal and dispersion derivatives

none for no hrf convolution 

#model_covariates: additional control covariates

list of additional covariates ; possible values:  (default: motion / art)

motion : (used in combination with functional_realignment preprocessing step) adds 6 motion parameters

motion+deriv: adds 12 motion parameters (6 motion parameters + their first-order derivatives)

motion+deriv+square: adds 24 motion parameters (6 motion parameters + their first-order derivatives + both squared)

art: (used in combination with functional_art preprocessing step) ads scrubbing covariates (one per outlier scan)

linear: adds a linear detrending term (within each run/session)

denoise: (used in combination with functional_regression step) adds covariates defined or estimated in functional_regression step

<covariatename>: (any existing 1st-level covariate) adds manually-defined covariates

<filename>.mat (one file per run) : adds manually-defined covariates

#model_serial    : serial correlation

none / AR(1) : serial correlation modeling (default: AR(1))

#model_session   : random/fixed task*session effects

1/0 : estimates session-specific task effects (default value: 1; when set to 0 SPM estimates session-invariant task effects -it assumes same/constant effect across all sessions-; alternatively set to cell array of task names in order to specify individual task effects that will be model as session-invariant)

#model_folder    : output folder

folder where model_name subfolder will be created (default '../'; relative to the <pipelineID> directory) 

set to the keyword 'root' to indicate the above location (root/<pipelineID>)

set to the keyword 'preproc' to indicate the root/<pipelineID>/results/firstlevel location

#mthresh         : implicit mask

implicit masking threshold as proportion of global signal (default value: 0.8)

#explicitmask    : explicit mask 

optional explicit mask (filename)

#hpf             : high-pass filter

high-pass filter threshold -in seconds- (default value: 128s)

#lpf             : low-pass filter

low-pass filter threshold -in seconds- (default value: 0s = no low-pass filter)

#contrast_addsession: adds session-specific task contrasts

0/1 for each contrast defined in #contrasts field above, create additional SESSION##_[contrastname] including only the n-th run data (default 0)

#contrast_addcv     : adds across-sessions cross-validation task contrasts

 0/1 for each contrast defined in #contrasts field above, create additional SESSION##_[contrastname] and ORTH_TO_SESSION##_[contrastname] including only the n-th run and all but the n-th run data, respectively (default 0)

#contrast_addoddeven: adds odd/even session task contrast

0/1 for each contrast defined in #contrasts field above, create additional ODD_[contrastname] and EVEN_[contrastname] including only odd and even numbered runs, respectively (default 0)

#contrast_removenonestimablecols: disregard nonestimable regressors

0/1 When defining contrast vectors skip design-matrix columns that are not fully estimable individually (default 0)

#contrast_removenonestimablecontrasts: disregard nonestimable contrasts

0/1 Skip contrasts that are not fully estimable (default 1)

#contrast_removeexistingcontrasts: remove previous contrasts

0/1 removes/deletes any older/existing contrasts in this first-level analysis (default 1)

#contrast_addevent: adds event-specific task contrasts

(for non-parametric modulations) 0/1 for each contrast defined in #contrasts field above, automatically create additional [contrastname]_EVENT## contrasts for each individual condition (default 0)

#contrastonly    : skip model definition and estimation

1/0 : skips first-level model definition/estimation step and performs only contrast definition/estimation [0]

pipeline_model_Default.cfg example:

#functional_label 

minimallysmoothed

#model_basis

hrf+deriv

#model_covariates

denoise

#model_serial

AR(1)

#hpf

128