Regressor has extra item, causing mismatch

[smeisler]

Environment

• fitlins version: 0.11
• Running via singularity on linux HPC

Expected and Observed Behavior

I had run the same basic model (subject-level contrast maps) for different tasks in the same subjects, and it ran fine. The only difference between the models are the trial_types and resulting contrast names.

I get the following error:

Traceback (most recent call last):
  File "/opt/miniconda-latest/envs/neuro/lib/python3.9/site-packages/nipype/pipeline/plugins/multiproc.py", line 67, in run_node
    result["result"] = node.run(updatehash=updatehash)
  File "/opt/miniconda-latest/envs/neuro/lib/python3.9/site-packages/nipype/pipeline/engine/nodes.py", line 527, in run
    result = self._run_interface(execute=True)
  File "/opt/miniconda-latest/envs/neuro/lib/python3.9/site-packages/nipype/pipeline/engine/nodes.py", line 645, in _run_interface
    return self._run_command(execute)
  File "/opt/miniconda-latest/envs/neuro/lib/python3.9/site-packages/nipype/pipeline/engine/nodes.py", line 771, in _run_command
    raise NodeExecutionError(msg)
nipype.pipeline.engine.nodes.NodeExecutionError: Exception raised while executing Node _design_matrix0.

Traceback:
	Traceback (most recent call last):
	  File "/opt/miniconda-latest/envs/neuro/lib/python3.9/site-packages/nipype/interfaces/base/core.py", line 398, in run
	    runtime = self._run_interface(runtime)
	  File "/opt/miniconda-latest/envs/neuro/lib/python3.9/site-packages/fitlins/interfaces/nistats.py", line 124, in _run_interface
	    mat = dm.make_first_level_design_matrix(
	  File "/opt/miniconda-latest/envs/neuro/lib/python3.9/site-packages/nilearn/glm/first_level/design_matrix.py", line 337, in make_first_level_design_matrix
	    assert add_regs_.shape[0] == np.size(frame_times), ValueError(
	AssertionError: Incorrect specification of additional regressors: length of regressors provided: 425, number of time-frames: 424

The number of time points in the run is indeed 424, and the confounds.tsv file also as 424 rows (not including the header column), so I am not seeing where this extra column is coming from.

I used the same fMRIPrep version to produce the different tasks, and they completed without errors. So it is unclear why this one task is causing problems while the others are not.

Steps to Reproduce

1. run fMRIPrep 22.0.0 with --cifti-output 91k flag
2. run fitlins to process cifti data: --space "fsLR" --drop-missing -s 5 --desc-label "" flags. Error also happens with volumetric data.
3. Model is below:

{
   "Name":"face-model-6MP5ACompCor",
   "BIDSModelVersion":"1.0.0",
   "Description":"",
   "Input":{
      "task":["face"]
   },
   "Nodes":[
      {
         "Level":"run",
         "Name":"runFace6MP5ACompCor",
         "GroupBy":[
            "subject",
	    "session",
	    "run"
         ],
	"Transformations":{
            "Transformer":"pybids-transforms-v1",
            "Instructions":[
               {
                  "Name":"Factor",
                  "Input":[
                     "trial_type"
                  ]
               },
               {
                  "Name":"Convolve",
                  "Input":[
                     "trial_type.Happy",
                     "trial_type.Fearful",
		     "trial_type.Sad",
                     "trial_type.Neutral",
                     "trial_type.Object"
                  ],
                  "Model":"spm"
               }
            ]
         },
         "Model":{
            "X":[
               "trial_type.Happy",
               "trial_type.Fearful",
	       "trial_type.Sad",
               "trial_type.Neutral",
               "trial_type.Object",
	       "non_steady_state*",
	       "trans_x",
               "trans_y",
               "trans_z",
               "rot_x",
               "rot_y",
               "rot_z",
	       "a_comp_cor_00",
	       "a_comp_cor_01",
	       "a_comp_cor_02",
	       "a_comp_cor_03",
	       "a_comp_cor_04",
	       "cosine*",
               1
            ],
           "Type":"glm"
         },
         "Contrasts":[
            {
               "Name":"faces_gt_object",
               "ConditionList":[
                  "trial_type.Happy",
                  "trial_type.Fearful",
	          "trial_type.Sad",
                  "trial_type.Neutral",
                  "trial_type.Object"
               ],
               "Weights":[
                  0.25,
		  0.25,
		  0.25,
		  0.25,
                  -1
               ],
               "Test":"t"
            },
	    {
               "Name":"sad_gt_neutral",
               "ConditionList":[
	          "trial_type.Sad",
                  "trial_type.Neutral"
               ],
               "Weights":[
                  1,
                  -1
               ],
               "Test":"t"
            },
	    {
               "Name":"fearful_gt_neutral",
               "ConditionList":[
	          "trial_type.Fearful",
                  "trial_type.Neutral"
               ],
               "Weights":[
                  1,
                  -1
               ],
               "Test":"t"
            },
	    {
               "Name":"happy_gt_neutral",
               "ConditionList":[
	          "trial_type.Happy",
                  "trial_type.Neutral"
               ],
               "Weights":[
                  1,
                  -1
               ],
               "Test":"t"
            }
         ]
      },
     {
    "Level":"session",
	"Name":"sessionFace6MP5ACompCor",
	"GroupBy":["session",
		"subject",
		"contrast"],
	"Model":{
            "X":[1],
            "Type":"meta"
            },
	"DummyContrasts": {"Test": "t"}
     	}
   ],
   "Edges":[
	{
      "Source": "runFace6MP5ACompCor",
      "Destination": "sessionFace6MP5ACompCor"
    }
   ]
}

4. Events file:

[effigies]

Do you have a pklz crash file? If we can use nipypecli to rerun the node and drop into the debugger, that would be simplest.

[smeisler]

Here is what I have in the working directory that corresponds to the crashing node:

/om2/scratch/Sun/smeisler/INSIDE_fmriprep/sub-inside7002/fitlins_wf/design_matrix/
├── _0x244397bf1abb9d04b3ef1141425a3d77.json
├── _inputs.pklz
├── mapflow
│   ├── _design_matrix0
│   │   ├── _inputs.pklz
│   │   ├── _node.pklz
│   │   ├── _report
│   │   │   └── report.rst
│   │   └── result__design_matrix0.pklz
│   └── _design_matrix1
│       ├── _inputs.pklz
│       ├── _node.pklz
│       ├── _report
│       │   └── report.rst
│       └── result__design_matrix1.pklz
├── _node.pklz
├── _report
│   └── report.rst
└── result_design_matrix.pklz

[effigies]

I think we emit crashfiles, either in the working directory or the scratch directory. They would start with crash and either be .txt or .pklz files.

[effigies]

Yes, it should be /om2/scratch/Sun/smeisler/INSIDE_fmriprep/sub-inside7002/crash*.

[smeisler]

Yes I found the crash files there.

[effigies]

If it's a .pklz file, run nipypecli crash -r -d <crashfile>, you'll drop into a pdb shell. (You might need to do singularity exec ... nipypecli ....)

From there you can go up a level and see the variables that are present here:

fitlins/fitlins/interfaces/nistats.py

Lines 124 to 130 in 222d9ab

	mat = dm.make_first_level_design_matrix(
	frame_times=np.arange(vols) * info['repetition_time'],
	add_regs=dense,
	hrf_model=None, # XXX: Consider making an input spec parameter
	add_reg_names=column_names,
	drift_model=drift_model,
	)

So you'd want to see what dense looks like, since that will have 425 rows.

[smeisler]

The crash files were .txt. dense is a .h5 file called /om2/scratch/Sun/smeisler/INSIDE_fmriprep/sub-inside7002/fitlins_wf/loader/run/run-1_subject-inside7002_dense.h5 (or run-2 depending on the run)
I opened it in a h5 viewer, but it is not clear where a table is encoded.

[effigies]

Ah, I'm sorry I dropped this. Apologies, the way to look at dense is with Pandas:

fitlins/fitlins/interfaces/nistats.py

Line 92 in 63fcc78

dense = pd.read_hdf(info['dense'], key='dense')

So:

>>> pd.read_hdf('/om2/scratch/Sun/smeisler/INSIDE_fmriprep/sub-inside7002/fitlins_wf/loader/run/run-1_subject-inside7002_dense.h5', key='dense')