stats fmriprep aroma #991

Rageve · 2023-04-27T10:21:17Z

Rageve
Apr 27, 2023

Dear Rami,

now I'm starting to run the statistics using my own dataset.
We conducted a food-cue reactivity task on two sessions and with two runs each. Preprocessing with fmriprep worked fine. We selected the flag --use-aroma to nonaggressively correct movement related noise components. In this case fmriprep creates a file sub-1085_ses-V0_task-food_run-1_space-MNI152NLin6Asym_desc-smoothAROMAnonaggr_bold.nii that is already smoothed with FWHM of 6mm.
The preprocessed data is now in the folder

.../BIDS/Tuebingen/inputs/fmriprep/sub-1085/ses-V0/func/
.../BIDS/Tuebingen/inputs/fmriprep/sub-1085/ses-V9/func

In your default setup you expect smoothing, because fmriprep is normally not smoothing the data during preprocessing. To become more familiar with the stats processing I created a simple default model

task_in_the_model ={'food'}

bidspm(bids_dir, output_dir, 'dataset', ...
        'action', 'default_model', ...
        'task', task_in_the_model, ...
        'space', {'MNI152NLin6Asym'});

the model was then created and saved at /server/fo2-17/data/NASAL_2018/DAGen_Bromo/BIDS/Tuebingen/outputs/derivatives/models/model-defaultFood_smdl.json
It's only a test file and there are some modifications in the design matrix (i.e. confounds) and contrast settings necessary (I use visual studio code for editing).

then I tried to run the stats using

bidspm(bids_dir, output_dir, 'subject', ...
         'participant_label', {subject_label}, ...
         'action', 'stats', ...
         'preproc_dir', prepro_dir, ...
         'task', {'food'}, ...
         'model_file', model_file);

and the script stopped with the error message

Indexing subject: sub-1085 [....]

[10:22:13] bidspm - INFO				getData
WILL WORK ON SUBJECTS
	- 1085


[10:22:13] bidspm - INFO				printWorkflowName
WORKFLOW: SUBJECT LEVEL GLM

[10:22:13] bidspm - INFO				bidsFFX

 PROCESSING NODE: run

Warning:
[10:22:13] bidspm - WARNING				bidsFFX
No data found in /server/fo2-17/data/NASAL_2018/DAGen_Bromo/BIDS/Tuebingen/inputs/fmriprep
for subject 1085 for filter:

	- desc: {smth6}
	- extension: {.nii.*}
	- modality: {func}
	- prefix: {}
	- space: {MNI152NLin6Asym}
	- sub: {^1085$}
	- suffix: {bold}
	- task: {food}

Available spaces are:

	- MNI152NLin2009cAsym
	- MNI152NLin6Asym

Available tasks are:

and ...

Now there are several question. Should I copy the data to a folder outside fmriprep (bidspm-preproc) and how? How can I change the input of
the - desc flag {smth6} to {smooth}. How can I adapt the script that only one session (ses-V0) will be analyzed?

Best
Ralf
model-defaultFood_smdl.zip

Remi-Gau · 2023-04-27T10:47:10Z

Remi-Gau
Apr 27, 2023
Maintainer

OK if you want to use fmriprep data without extra smoothing, you must first indeed copy the data you need in a bidspm-preproc folder.

For this you can use the copy action

https://bidspm.readthedocs.io/en/latest/usage_notes.html

task_in_the_model ={'food'}

bidspm(bids_dir, output_dir, 'subject', ...
        'action', 'copy', ...
        'participant_label', {subject_label}, ...
        'bids_filter_file', struct([]), ...
        'verbosity', 2, ...
        'space', {'MNI152NLin6Asym'}, ...
        'options', struct([]), ...
        'task', task_in_the_model, ...
        'skip_validation', true)

You can use the 'bids_filter_file', struct([]) to specify more filter for what data you want to copy.

The default structure used is this one:

struct('fmap', struct('modality', 'fmap'), ...
       'bold', struct('modality', 'func', 'suffix', 'bold'), ...
       't2w',  struct('modality', 'anat', 'suffix', 'T2w'), ...
       't1w',  struct('modality', 'anat', 'space', '', 'suffix', 'T1w'), ...
       'roi',  struct('modality', 'roi', 'suffix', 'mask'), ...
       'xfm',  struct('modality', 'anat', 'suffix', 'xfm', 'to', 'T1w'));

https://bidspm.readthedocs.io/en/latest/dev_doc.html#src.defaults.checkOptions

So in your case you could do:

bids_filter_file = struct('fmap', struct('modality', 'fmap'), ...
       'bold', struct('modality', 'func', 'suffix', 'bold', 'desc', 'smoothAROMAnonaggr'), ... % < ---- specify the 'desc' here
       't2w',  struct('modality', 'anat', 'suffix', 'T2w'), ...
       't1w',  struct('modality', 'anat', 'space', '', 'suffix', 'T1w'), ...
       'roi',  struct('modality', 'roi', 'suffix', 'mask'), ...
       'xfm',  struct('modality', 'anat', 'suffix', 'xfm', 'to', 'T1w'));


task_in_the_model ={'food'}

bidspm(bids_dir, output_dir, 'subject', ...
        'action', 'copy', ...
        'participant_label', {subject_label}, ...
        'bids_filter_file', bids_filter_file, ...
        'verbosity', 2, ...
        'space', {'MNI152NLin6Asym'}, ...
        'options', struct([]), ...
        'task', task_in_the_model, ...
        'skip_validation', true)

1 reply

Remi-Gau Apr 27, 2023
Maintainer

Note that this bids filter file is the same strategy used by fmriprep: https://fmriprep.org/en/stable/faq.html#how-do-i-select-only-certain-files-to-be-input-to-fmriprep

So you can also use a json file to define those filters:

>> bids_filter_file = struct('fmap', struct('modality', 'fmap'), ...
       'bold', struct('modality', 'func', 'suffix', 'bold', 'desc', 'smoothAROMAnonaggr'), ...
       't2w',  struct('modality', 'anat', 'suffix', 'T2w'), ...
       't1w',  struct('modality', 'anat', 'space', '', 'suffix', 'T1w'), ...
       'roi',  struct('modality', 'roi', 'suffix', 'mask'), ...
       'xfm',  struct('modality', 'anat', 'suffix', 'xfm', 'to', 'T1w'));
>> bids.util.jsonwrite(fullfile(pwd, 'bids_filter_file.json'), bids_filter_file)

This will create this file:

{
  "fmap": {
    "modality": "fmap"
  },
  "bold": {
    "modality": "func",
    "suffix": "bold",
    "desc": "smoothAROMAnonaggr"
  },
  "t2w": {
    "modality": "anat",
    "suffix": "T2w"
  },
  "t1w": {
    "modality": "anat",
    "space": "",
    "suffix": "T1w"
  },
  "roi": {
    "modality": "roi",
    "suffix": "mask"
  },
  "xfm": {
    "modality": "anat",
    "suffix": "xfm",
    "to": "T1w"
  }
}

that you should be able to use like this:

task_in_the_model ={'food'}

bidspm(bids_dir, output_dir, 'subject', ...
        'action', 'copy', ...
        'participant_label', {subject_label}, ...
        'bids_filter_file', fullfile(pwd, 'bids_filter_file.json'), ...
        'verbosity', 2, ...
        'space', {'MNI152NLin6Asym'}, ...
        'options', struct([]), ...
        'task', task_in_the_model, ...
        'skip_validation', true)

Remi-Gau · 2023-04-27T10:55:31Z

Remi-Gau
Apr 27, 2023
Maintainer

Your bids stats model looks mostly fine:

I removed:

transformation
the contrast
and modified the GroupBy at the dataset level to for each contrast from the subject level.

{
  "Name": "default_food_model",
  "BIDSModelVersion": "1.0.0",
  "Description": "default BIDS stats model for food task",
  "Input": {
    "task": ["food"],
    "space": ["MNI152NLin6Asym"]
  },
  "Nodes": [
    {
      "Level": "Run",
      "Name": "run",
      "GroupBy": [
        "run",
        "session",
        "subject"
      ],
      "Model": {
        "X": [
          "trial_type.hi_savory",
          "trial_type.hi_sweet",
          "trial_type.lo_savory",
          "trial_type.lo_sweet",
          "1",
          "trans_?",
          "rot_?",
          "non_steady_state_outlier*",
          "motion_outlier*"
        ],
        "Type": "glm",
        "HRF": {
          "Variables": [
            "trial_type.hi_savory",
            "trial_type.hi_sweet",
            "trial_type.lo_savory",
            "trial_type.lo_sweet"
          ],
          "Model": "spm"
        },
        "Options": {
          "HighPassFilterCutoffHz": 0.008,
          "Mask": {
            "desc": ["brain"],
            "suffix": ["mask"]
          }
        },
        "Software": {
          "SPM": {
            "SerialCorrelation": "FAST",
            "InclusiveMaskingThreshold": 0.8
          }
        }
      },
      "DummyContrasts": {
        "Test": "t",
        "Contrasts": [
          "trial_type.hi_savory",
          "trial_type.hi_sweet",
          "trial_type.lo_savory",
          "trial_type.lo_sweet"
        ]
      }
    },
    {
      "Level": "Subject",
      "Name": "subject",
      "GroupBy": [
        "subject",
        "contrast"
      ],
      "Model": {
        "X": [1],
        "Type": "glm"
      },
      "DummyContrasts": {
        "Test": "t"
      }
    },
    {
      "Level": "Dataset",
      "Name": "dataset",
      "GroupBy": ["contrast"],
      "Model": {
        "X": [1],
        "Type": "glm"
      },
      "DummyContrasts": {
        "Test": "t"
      }
    }
  ],
  "Edges": [
    {
      "Source": "run",
      "Destination": "subject"
    },
    {
      "Source": "subject",
      "Destination": "dataset"
    }
  ]
}

0 replies

Rageve · 2023-04-27T15:06:40Z

Rageve
Apr 27, 2023
Author

When I try to copy using your suggestions I receive the error message ```matlab 15:19:43] bidspm - INFO getData WILL WORK ON SUBJECTS - 1085 [15:19:43] bidspm - INFO No data found for this query: - desc: {preproc, brain} - modality: {anat} - space: {MNI152NLin6Asym} - suffix: {T1w, bold, mask} - sub: {1085} in dataset: /server/fo2-17/data/NASAL_2018/DAGen_Bromo/BIDS/Tuebingen/inputs/fmriprep ``` The data is actually in the dataset folder but with different sessions .../fmriprep/sub-1085/ses-V0/func .../fmriprep/sub-1085/ses-V9/func .../fmriprep/sub-1085/anat anat is in a superior folder because fmriprep is searching for all T1-images in a certain subject and create one anat folder below the subject folder. In the other .../sub-1085/ses-V0/anat folder there is only a file sub-1085_ses-V0_from-orig_to-T1w_mode-image_xfm.txt with affine transformation parameters. Should I define the two sessions (and runs) in the bids_filter_file? How to handle the different level of the anat folder? I just realized that in the anat the data has a different space (MNI152NLin2009cAsym - but the default output for anat processing in fmriprep). Defining both spaces or the 2009 space in the structure for space does not change the error.

4 replies

Remi-Gau Apr 27, 2023
Maintainer

Ah this is one case I have not yet encountered yet.

I think this is related to something that still needs to be implemented in bids matlab.

bids-standard/bids-matlab#493

Remi-Gau Apr 27, 2023
Maintainer

Until I improve bids matlab to handle this
I think that you should be able to specify which modality to copy by creating an option structure.

Options = struct('query', struct('modality', {'func'}));

You can then pass this structure in your bidspm command.

Remi-Gau Apr 27, 2023
Maintainer

bidspm(bids_dir, output_dir, 'subject', ...
        'action', 'copy', ...
        'participant_label', {subject_label}, ...
        'bids_filter_file', bids_filter_file, ...
        'verbosity', 2, ...
        'space', {'MNI152NLin6Asym'}, ...
        'options', Options, ... %. <------ see here
        'task', task_in_the_model, ...
        'skip_validation', true)

Remi-Gau Apr 27, 2023
Maintainer

Something I can also implement fairly quickly is for the copy action to throw warnings and not errors when no file is found to copy.

Rageve · 2023-04-28T07:39:24Z

Rageve
Apr 28, 2023
Author

using the options flag, the copying works with the defaults files (including the confounds_timeseries.tsv files) ``` copying /server/fo2-17/data/NASAL_2018/DAGen_Bromo/BIDS/Tuebingen/inputs/fmriprep/sub-1085/ses-V0/func/sub-1085_ses-V0_task-food_run-1_space-MNI152NLin2009cAsym_desc-brain_mask.nii.gz --> /server/fo2-17/data/NASAL_2018/DAGen_Bromo/BIDS/Tuebingen/outputs/derivatives/bidspm-preproc/derivatives/bidspm-preproc/../bidspm-preproc/sub-1085/ses-V0/func/sub-1085_ses-V0_task-food_run-1_space-MNI152NLin2009cAsym_desc-brain_mask.nii.gz ``` but not with the already smoothed files that were stated in in the bids_filter_file ``` bids_filter_file = struct('fmap', struct('modality', 'fmap'), ... 'bold', struct('modality', 'func', 'suffix', 'bold', 'desc', 'smoothAROMAnonaggr'), ... 't2w', struct('modality', 'anat', 'suffix', 'T2w'), ... 't1w', struct('modality', 'anat', 'space', 'MNI152NLin2009cAsym', 'suffix', 'T1w'), ... 'roi', struct('modality', 'roi', 'suffix', 'mask'), ... 'xfm', struct('modality', 'anat', 'suffix', 'xfm', 'to', 'T1w')); ``` should I modify the options to select the specific 'bold' files

0 replies

Rageve · 2023-04-28T09:19:57Z

Rageve
Apr 28, 2023
Author

The bids_filter_file shows the correct field ``` bids_filter_file.bold ans = struct with fields: modality: 'func' space: 'MNI152NLin6Asym' suffix: 'bold' desc: 'smoothAROMAnonaggr' ``` but when running a different desc reference is listed ``` No data found for this query: - desc: {preproc, brain} - modality: {func} - space: {MNI152NLin6Asym} - suffix: {T1w, bold, mask} - sub: {1424} - task: {food} ```

1 reply

Remi-Gau Apr 28, 2023
Maintainer

OK then this is a bug and I need to fix this.

Will open a separate issue to track this.

Remi-Gau · 2023-05-09T00:49:28Z

Remi-Gau
May 9, 2023
Maintainer

OK so this problem should be fixed.

you should be able to copy smoothAROMAnonaggr files by specifying it in the bids_filter_json

I have a test where it works:

bidspm/tests/tests_cli/test_bidspm_copy.m

Lines 60 to 71 in e684f7a

    
           %% now with more complex filter as struct 
        
           bids_filter_file = struct('bold', struct('modality', 'func', ... 
        
                                                    'suffix', 'bold', ... 
        
                                                    'desc', {'smoothAROMAnonaggr'; ... 
        
                                                             'preproc'})); 
        
           bidspm(inputPath, outputPath, 'subject', ... 
        
                  'action', 'copy', ... 
        
                  'task', {'auditory'}, ... 
        
                  'space', {'MNI152NLin6Asym', 'T1w'}, ... 
        
                  'bids_filter_file', bids_filter_file, ... 
        
                  'verbosity', 0);

you also should not have to specify opt.query.modality = {'func'} to avoid crashes when the anat cannot be copied, the copying process has been made tolerant to some errors.

so I am going to close this discussion, but feel free to reopen it or open a new one or an issue if you still encounter problems with this.

0 replies

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

stats fmriprep aroma #991

{{title}}

{{editor}}'s edit

{{editor}}'s edit

Replies: 6 comments 6 replies

{{title}}

{{title}}

{{title}}

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{title}}

Select a reply

stats fmriprep aroma #991

Rageve Apr 27, 2023

Replies: 6 comments · 6 replies

Remi-Gau Apr 27, 2023 Maintainer

Remi-Gau Apr 27, 2023 Maintainer

Remi-Gau Apr 27, 2023 Maintainer

Rageve Apr 27, 2023 Author

Remi-Gau Apr 27, 2023 Maintainer

Remi-Gau Apr 27, 2023 Maintainer

Remi-Gau Apr 27, 2023 Maintainer

Remi-Gau Apr 27, 2023 Maintainer

Rageve Apr 28, 2023 Author

Rageve Apr 28, 2023 Author

Remi-Gau Apr 28, 2023 Maintainer

Remi-Gau May 9, 2023 Maintainer

Rageve
Apr 27, 2023

Replies: 6 comments 6 replies

Remi-Gau
Apr 27, 2023
Maintainer

Remi-Gau Apr 27, 2023
Maintainer

Remi-Gau
Apr 27, 2023
Maintainer

Rageve
Apr 27, 2023
Author

Remi-Gau Apr 27, 2023
Maintainer

Remi-Gau Apr 27, 2023
Maintainer

Remi-Gau Apr 27, 2023
Maintainer

Remi-Gau Apr 27, 2023
Maintainer

Rageve
Apr 28, 2023
Author

Rageve
Apr 28, 2023
Author

Remi-Gau Apr 28, 2023
Maintainer

Remi-Gau
May 9, 2023
Maintainer