[ENH] add roi based GLM using mars bar #340

Remi-Gau · 2021-04-03T09:20:08Z

Other changes:

change the option to skip GLM QA:

it is now opt.glm.QA.do

it is possible to decide where the GLM output will go by using opt.dir.stats. The default is now:

cpp_spm-stats

here is what the output structure could look like

demos/spm_face_rep/
├── cfg
├── inputs
│   └── source
├── models
│   ├── model-defaultFacerepetition_smdl.json
│   ├── model-faceRepetitionResolution15_smdl.json
│   ├── model-faceRepetitionResolution1_smdl.json
│   ├── model-faceRepetitionResolution2_smdl.json
│   └── model-faceRepetition_smdl.json
└── outputs
    ├── derivatives
    │   ├── cpp_spm
    │   │   ├── JOBS
    │   │   │   └── facerepetition
    │   │   │       └── sub-01
    │   │   └── sub-01
    │   │       ├── anat
    │   │       ├── figures
    │   │       ├── func
    │   │       └── stats     <---------- still possible to keep the stats with the preprocessing by using `opt.dir.stats = opt.derivatesDir`
    │   │           └── task-facerepetition_space-MNI_FWHM-6
    │   ├── cpp_spm-res1   <--------- different preprocessing have to be kept in different subfolders
    │   │   ├── JOBS
    │   │   │   └── facerepetition
    │   │   │       └── sub-01
    │   │   └── sub-01
    │   │       ├── anat
    │   │       ├── figures
    │   │       └── func
    │   ├── cpp_spm-res1.5
    │   │   ├── JOBS
    │   │   │   └── facerepetition
    │   │   │       └── sub-01
    │   │   └── sub-01
    │   │       ├── anat
    │   │       ├── figures
    │   │       └── func
    │   ├── cpp_spm-res2
    │   │   ├── JOBS
    │   │   │   └── facerepetition
    │   │   │       └── sub-01
    │   │   └── sub-01
    │   │       ├── anat
    │   │       ├── figures
    │   │       └── func
    │   └── cpp_spm-stats   <----- all the stats can be centralized
    │       ├── JOBS
    │       │   └── facerepetition
    │       │       └── sub-01
    │       └── sub-01
    │           └── stats
    │               ├── task-facerepetition_space-MNI_FWHM-6_desc-resamplingResolution1
    │               ├── task-facerepetition_space-MNI_FWHM-6_desc-resamplingResolution15
    │               └── task-facerepetition_space-MNI_FWHM-6_desc-resamplingResolution2
    └── raw

codecov · 2021-04-03T12:00:59Z

Codecov Report

Merging #340 (d3f0a1b) into dev (c6b1b96) will decrease coverage by 1.98%.
The diff coverage is 12.98%.

@@            Coverage Diff             @@
##              dev     #340      +/-   ##
==========================================
- Coverage   63.33%   61.35%   -1.99%     
==========================================
  Files         111      114       +3     
  Lines        1762     1832      +70     
==========================================
+ Hits         1116     1124       +8     
- Misses        646      708      +62

Impacted Files	Coverage Δ
src/batches/setBatchEstimateModel.m	`0.00% <0.00%> (ø)`
src/group_level/getRFXdir.m	`100.00% <ø> (ø)`
src/templates/bidsTemplateWorkflow.m	`0.00% <0.00%> (ø)`
src/workflows/bidsCreateROI.m	`0.00% <0.00%> (ø)`
src/workflows/bidsFFX.m	`0.00% <0.00%> (ø)`
src/workflows/bidsRFX.m	`0.00% <0.00%> (ø)`
src/workflows/bidsRoiBasedGLM.m	`0.00% <0.00%> (ø)`
src/subject_level/getFFXdir.m	`87.50% <66.66%> (-12.50%)`	⬇️
src/defaults/checkOptions.m	`100.00% <100.00%> (ø)`
src/setStatsDir.m	`100.00% <100.00%> (ø)`
... and 3 more

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c6b1b96...d3f0a1b. Read the comment docs.

Remi-Gau · 2021-04-03T14:37:04Z

@marcobarilari
@CerenB

maybe the options mentioned above might be of interest to you.

CerenB

nice! I left some questions to get confirmation from you. And I was wondering this structure of the folders will be applied to the whole-brain analysis, it is not only for ROIs, right?
e.g. if I opt to use no ROIs, and change the resolution in the analysis, I can re-run the spatial preprocessing script and the pipeline would create a new folder (actually new folder structure)

CerenB · 2021-04-07T10:12:48Z

demos/spm_face_rep/face_rep_roi_analysis.m

+
 bidsCreateROI(opt);
+
+opt.glm.roibased.do = true;


if this option is set to false, would it do the whole brain analysis with different resolution spatial preprocessing?

CerenB · 2021-04-07T10:14:19Z

demos/spm_face_rep/face_rep_resolution.m

+  opt.model.file = newModel;
+
+  spm_jsonwrite(newModel, content, struct('indent', '   '));
+


this is only for this function to create names (in order to do so, we need to create different model.json files) for different resolution? so that spatial preprocssing folders would be different?

Yeah this is only to create a different GLM folder name.

This is a bit "ugly" because everything in those json is the same: but I was trying to find a way to create different folder names that would not involve adding another option to the pile and would instead reuse something we already have to create GLM folder names.

Does that make sense?

Another option could have been to have a different cpp_spm-stats folder for each resolution but then all the stats would not be centralized.

I think there is only so much data curation that can be automatized.

CerenB · 2021-04-07T10:16:03Z

src/subject_level/getFFXdir.m

                    'stats', ...
                    glmDirName);

+  if opt.glm.roibased.do


Yeah that's pretty much it: one of the few thing this flag does is append a suffix to the GLM folder.

This way in this case you can use the same model json for both whole brain and roi based analysis

CerenB · 2021-04-07T10:23:08Z

src/workflows/bidsRoiBasedGLM.m

@@ -0,0 +1,115 @@
+% (C) Copyright 2021 CPP BIDS SPM-pipeline developers
+
+function bidsRoiBasedGLM(opt)


this function is for running GLM on only ROIs, right? instead of whole brain. Here is a irrelevant question!
I usually do not prefer roi-based GLM, and go with whole brain. Then extract the betas from a given ROI. Do you know which functions could be useful for extracting betas in a given ROI in cpp-spm?

spm_summarize ?

Remi-Gau

And I was wondering this structure of the folders will be applied to the whole-brain analysis, it is not only for ROIs, right?

Yes that's correct

Remi-Gau · 2021-04-07T11:06:31Z

I was just thinking.

Maybe we should add some "child protection": if opt.glm.roibased.do is true then bidsFFX should NOT run and bidsRoiBasedGLM should.

And then the other way around when set to false.

What do you think?

Maybe we should throw an error with a message saying what the expected option setting should be for each workflow to run.

… wrong options AKA child protection

Remi-Gau · 2021-04-07T16:05:28Z

One thing that will need to be done in another PR is the ability to compute contrasts from the bids model and the content of the options.

At the moment the roi based workflow only does it assuming for all "events" taken together.

Remi-Gau added 6 commits April 3, 2021 10:31

set up roi based GLMs

6b5cac3

add mars bar roi based GLM estimation

9bf5906

fit events and compute percent signal change

2c88cfd

refactor MoAE to allow running with octave

382b8b7

update tests

b599754

fix test

0eef7e1

Remi-Gau added 3 commits April 3, 2021 15:44

update tests

0c1ef80

update face rep demo

c57312c

reset demos

bce592c

Remi-Gau linked an issue Apr 3, 2021 that may be closed by this pull request

organize output of FFX and RFX in separate derivatives folder #96

Closed

CerenB approved these changes Apr 7, 2021

View reviewed changes

Remi-Gau commented Apr 7, 2021

View reviewed changes

add errors in FFX and roi based GLMs to prevent running them with the…

d3f0a1b

… wrong options AKA child protection

Remi-Gau linked an issue Apr 7, 2021 that may be closed by this pull request

Add template file to create new workflows #317

Closed

Remi-Gau merged commit 6010230 into cpp-lln-lab:dev Apr 7, 2021

Remi-Gau deleted the remi-roi_based_glm branch April 7, 2021 16:12

		opt.model.file = newModel;

		spm_jsonwrite(newModel, content, struct('indent', ' '));

		@@ -0,0 +1,115 @@
		% (C) Copyright 2021 CPP BIDS SPM-pipeline developers

		function bidsRoiBasedGLM(opt)

[ENH] add roi based GLM using mars bar #340

[ENH] add roi based GLM using mars bar #340

Uh oh!

Conversation

Remi-Gau commented Apr 3, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

codecov bot commented Apr 3, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Remi-Gau commented Apr 3, 2021

Uh oh!

CerenB left a comment

Choose a reason for hiding this comment

Uh oh!

CerenB Apr 7, 2021

Choose a reason for hiding this comment

Uh oh!

CerenB Apr 7, 2021

Choose a reason for hiding this comment

Uh oh!

Remi-Gau Apr 7, 2021

Choose a reason for hiding this comment

Uh oh!

CerenB Apr 7, 2021

Choose a reason for hiding this comment

Uh oh!

Remi-Gau Apr 7, 2021

Choose a reason for hiding this comment

Uh oh!

CerenB Apr 7, 2021

Choose a reason for hiding this comment

Uh oh!

Remi-Gau Apr 7, 2021

Choose a reason for hiding this comment

Uh oh!

Remi-Gau left a comment

Choose a reason for hiding this comment

Uh oh!

Remi-Gau commented Apr 7, 2021

Uh oh!

Remi-Gau commented Apr 7, 2021

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Remi-Gau commented Apr 3, 2021 •

edited

Loading

codecov bot commented Apr 3, 2021 •

edited

Loading