[R5K7] Pipeline reproduction (SPM - raw)

[cmaumet]

Softwares

SPM12 v7487 ,\nMatlab R2017b (9.3.0.713579)

Input data

raw data

Additional context

see description below

List of tasks

Please tick the boxes below once the corresponding task is finished. 👍

• 👌 A maintainer of the project approved the issue, by assigning a 🏁status: ready for dev label to it.
• 🌳 Create a branch on your fork to start the reproduction.
• 🌅 Create a file team_{team_id}.py inside the narps_open/pipelines/ directory. You can use a file inside narps_open/pipelines/templates as a template if needed.
• 📥 Create a pull request as soon as you completed the previous task.
• 🧠 Write the code for the pipeline, using Nipype and the file architecture described in docs/pipelines.md.
• 📘 Make sure your code is documented enough.
• 🐍 Make sure your code is explicit and conforms with PEP8.
• 🔬 Create tests for your pipeline. You can use files in tests/pipelines/test_team_* as examples.
• 🔬 Make sure your code passes all the tests you created (see docs/testing.md).

---

NARPS team description : R5K7

General

• teamID : R5K7
• NV_collection_link : https://neurovault.org/collections/4950/
• results_comments : Hypothesis 9 as phrased here could - strictly - not be tested. Rather, it is assumed that hypo 9 referred to "greater positive parametric effect of loss in amygdala for equal range vs. equal indifference conditions."
• preregistered : No
• link_preregistration_form : NA
• regions_definition : anatomical definitions

• vmPFC: a combination of the following maps of the Jülich cytoarchitectonic atlas contained in the SPM Anatomy Toolbox (version 3), thresholded at a probability of P >= 0.2: OFC_Fo1, OFC_Fo2, FP2 und Cingul_s32 (left and right hemisphere each)
• amygdala: as above, with the following maps: SF, MF, IF, LB, CM (left and right each)
• ventral striatum: taken as is from the striatum atlas included in the FSL software package

• softwares : SPM12 v7487 ,
  Matlab R2017b (9.3.0.713579)
• general_comments : NA

Exclusions

• n_participants : 108
• exclusions_details : NA

Preprocessing

• used_fmriprep_data : No
• preprocessing_order : 1) motion correction

2. intersubject registration (normalization)
3. spatial smoothing

• brain_extraction : NA
• segmentation : NA
• slice_time_correction : NA
• motion_correction : SPM12, Realign & Unwarp using the phase map generated from the fieldmap data with SPM12 Fieldmap Toolbox v2.1 (default options).
  Other than defaults:

• estimation: quality 0.95, separation 3, two-pass procedure (i. registering to 1st scan, ii. registering to mean image), interpolation 7th degree B-spline;
• unwarp & reslice: interpolation 7th degree B-spline

• motion :
• gradient_distortion_correction : Also, the single-band reference EPI image was distortion-corrected using the SPM12 Fieldmap Toolbox v2.1.
• intra_subject_coreg : NA
• distortion_correction : NA
• inter_subject_reg : SPM12: For each run, the distortion-corrected single-band reference EPI image was co-registered to the mean EPI image obtained from Realignment & Unwarping using normalised mutual information. The single-band reference EPI image was then co-registered to the grey-matter probability map included in the Old Segmentation toolbox of SPM12 using normalised mutual information, with the mean EPI image and the distortion-corrected EPI time-series remaining aligned to the reference EPI image by applying the same shift parameteres. The single-band reference EPI image was subsequently normalized to SPM12's MNI152 template brain via the Unified Segmentation approach (Old Segment function in SPM12). To reduce the risk of overfitting we set the warp frequency cutoff to 45 limiting the discrete cosine transform bases, and set the sampling distance to 2. Beside the single-band reference EPI image, the resulting deformation field was also applied to the mean EPI image and the entire EPI time-series.
• intensity_correction : NA
• intensity_normalization : NA
• noise_removal : NA
• volume_censoring : NA
• spatial_smoothing : SPM12, fixed Gaussian kernel with 8 mm FWHM, performed in MNI152 space
• preprocessing_comments : NA

Analysis

• data_submitted_to_model : time series of 449 EPI volumes for each of 4 sessions in each of 108 participants
• spatial_region_modeled : NA
• independent_vars_first_level : - event-related design with each trial modelled with a duration of 4 sec and 3 linear parametric modulators (PMs orthogonalized via de-meaning against task and preceding PMs, respectively) for gain, loss and reaction time (in that order) as given in the .tsv log files

• canonical HRF plus temporal derivative
• 6 motion regressors (1st-order only) reflecting the 6 realignment parameters for translation and rotation movements obtained during preprocessing
• The above modelling was done for each session, with all 4 sessions being included in one 1st-level design matrix per participant.
• After model estimation, sum contrast images for each regressor of interest [task, gain (PM1), loss (PM2) and RT (PM3)] were computed across the 4 sessions in each participant.

• RT_modeling : pm
• movement_modeling : 1
• independent_vars_higher_level : A flexible factorial design was used to examine the effects of 4 factors of interest [task, gain (PM1), loss (PM2) and RT (PM3); cf. description above] for each of the 2 groups (Equal Indifference vs. Equal Range).

• condition effects per group modelled as factors (representing linear parametric modulation effects estimated at the 1st level)
• no covariates, no other between-group effects

• model_type : Mass Univariate
• model_settings : 1st-level model: "AR(1) + w" autocorrelation model in SPM, high-pass filter: 128 s
  2nd-level model: random-effects GLM implemented with weighted least squares (via SPM's restricted maximum likelihood estimation); both between-condition and between-group variances assumed to be unequal
• inference_contrast_effect : Linear T contrasts for the two parameters of interest (PM1 indicating linear hemodynamic changes with Gain value over trials within each subject, PM2 indicating such changes with Loss value) were used to test for the effects specified in the 9 hypotheses given.
• search_region : NA
• statistic_type : Voxel-wise
• pval_computation : standard parametric inference
• multiple_testing_correction : family-wise error correction, based on Random Field Theory
• comments_analysis : NA

Categorized for analysis

• region_definition_vmpfc : atlas Jülich cytoarchitectonic
• region_definition_striatum : atlas Oxford-Imanova Striatal Structural Atlas
• region_definition_amygdala : atlas Jülich cytoarchitectonic
• analysis_SW : SPM
• analysis_SW_with_version : SPM12
• smoothing_coef : 8
• testing : parametric
• testing_thresh :
• correction_method : GRTFWE voxelwise
• correction_thresh_ : p<0.05

Derived

• n_participants : 108
• excluded_participants : n/a
• func_fwhm : 8
• con_fwhm :

Comments

• excluded_from_narps_analysis : No
• exclusion_comment : N/A
• reproducibility : 2
• reproducibility_comment :

[cmaumet]

Note of things that we are unsure about:

Computation of vdm, we used:

• no masking
• -1 for blip direction (based on what was done for V55J)
• 29.15 for total readout time (based on V55J)
• non-EPI for EPI-based fieldmap (based on V55J)
• Yes for match VDM to EPI

Inter-subject registration:

• we assumed that the "distortion-corrected single-band reference EPI image" was the sbref image (i.e. we did not apply realign and unwarp to sbref)

[cmaumet]

(Note this reproductions is currently being developed as SPM (matlabbatch) code and will be converted to nipype in a second step). Corresponding file: https://github.com/cmaumet/narps_open_pipelines/blob/team_R5K7/narps_open/pipelines/matlabbatch_R5K7.m

[cmaumet]

Worked on this with @bclenet at the NOP hackathon Dec 12-13, 2024.

It looks like there is no nipype interface for SPM "flexible factorial" design (which is used at the second level): nipy/nipype#3619