I would like to run a first-level analysis on the HCP dataset using nipype + SPM. Some of you that are familiar with the data set know that each fMRI task was done both with an LR and RL phase encoding, so I took this into account in my SPM-analysis. I set up my workflow so that both runs are treated like two different sessions using concatenate_runs = False and by passing a list of two 4D nifti files to functional_runs and a list of two Bunch-objects to subject_info of SpecifySPMModel. The design matrix in general goes in the right direction (both runs are treated as different sessions), however I am getting these strange “phantom columns” (as I call them) that are just inversions to all my task conditions. So for example for my Sn(1)0back_body condition I get a second column next to it labeled Sn(1)0back_body(bf2), and so on for each condition in my design matrix. This leads also to twice the amount of beta-images when running EstimateModel afterwards.
This is how the input for SpecifySPMModel looks like:
{'concatenate_runs': False,
'output_units': 'secs',
'subject_info': [Bunch(conditions=['0bk_body',
'0bk_faces',
'0bk_places',
'0bk_tools',
'2bk_body',
'2bk_faces',
'2bk_places',
'2bk_tools'],
durations=[[27.5],
[27.5],
[27.5],
[27.5],
[27.5],
[27.5],
[27.5],
[27.5]],
onsets=[[36.118],
[221.84099999999998],
[250.05599999999998],
[107.435],
[150.51],
[79.207],
[178.605],
[7.996]]),
Bunch(conditions=['0bk_body',
'0bk_faces',
'0bk_places',
'0bk_tools',
'2bk_body',
'2bk_faces',
'2bk_places',
'2bk_tools'],
durations=[[27.5],
[27.5],
[27.5],
[27.5],
[27.5],
[27.5],
[27.5],
[27.5]],
onsets=[[107.448],
[36.145],
[150.537],
[221.96099999999998],
[7.996],
[178.739],
[250.042],
[79.35300000000001]])],
'bids_condition_column': 'trial_type',
'parameter_source': 'SPM',
'functional_runs': ['/output/workflow_files/nback_first_level_analysis/_subject_id_952863/model_specifier/ssub-952863_tfMRI_WM_LR.nii',
'/output/workflow_files/nback_first_level_analysis/_subject_id_952863/model_specifier/ssub-952863_tfMRI_WM_RL.nii'],
'input_units': 'secs',
'high_pass_filter_cutoff': 200.0,
'time_repetition': 0.72}
It seems that I am doing something wrong with defining my model, but I don’t know what I am doing wrong here? How can I set up my design matrix in such a way that I get a two session design with only one column for each condition (0back_tools, 0back_shapes, 0back_faces, 0back_places, 2back_tools, 2back_shapes, 2back_faces, 2back_places)?