It would be nice if unf2toad could check if the right files have been created at the end of the conversion.

Right now, if the user converts fieldmap images (phase and magnitude) and fails to specify the right directories, only one image is created with no error.

I put first level in the title since the verification is only on the filename. The second level could be on the file type and/or format itself.

What is the purpose of the white segmented image compute during the preprocessing task?

masking in nlmeans denoising step causes zero stripes in denoised data.
dwi mask should be warped from anatomical image as in fieldmap step.

Sometime Carollyn extract the B0 at postprocessing during the acquisition.
Evaluate if we should encourage and enforce this practice during a TOAD acquisition.

When folder has no DICOM within it, display multiple the folder name as choices.

eddy crash during correction task if the dwi and the b0's don't have the same voxel's size

• Instead of using our own code/converter of tractograms, we should use mrtrix' when it will be available.

Since vtk works on magma server. I suspect the exceptions raised by the wrapper fvtk instead of the librairies vtk itself

As discussed during the TOAD meeting we should add a new QA session where information about the different softwares version used during the pipeline. We should add also a summary of the analysis.

Christophe if you need help don't hesitate !

Someone should evaluate Beast as a replacement for FSL BET.
BET from FSL offers sometime disappointing results specifically when we expect bias field & neck cleanup.

BEAST: https://rpubs.com/conge/beast_intro
BET: http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/BET

ll est impératifs de créer des grilles (axes) sur les images d'assurance qualité afin d'avoir quelques repères visuels lors de leur consultation.

Some subjects fail to be well registered with anat. We should find a way to check if the registration went well and change parameters in case of "error".

1- Check registration
2- Add -usesqform in registration of magnitude, use norm.mgz instead of "anat freesurfer"

ASAP ;)

PS: What need to be modified:
-----> INFO: Coregistring magnitude image with the anatomical image produce by freesurfer

Denoising report should mention which denoising algorithm have been during the task.

We should think about a way to use TOAD for phantoms. Let's think about that for TOAD 4.0.

gifv format allows to save a lot of disk space when creating animated gif.
If you find a python library to do that, share it.

unf2toad have many flaw which make the tools hard to maintain.

• Add support for none Siemens dataset.

@cbedetti
@arnaudbore
Basile suggest an algorithm to detect (and possibly correct) spike in a diffusion image.

@cbedetti
freesurfer_lut options is use in many tasks but is declare into qa section.

You be good to create a new section for templates images

• We should suggest a serie of DICOM for conversion.
• We should check the serie of DICOM to make sure that it makes sense for conversion.
• Once you've done the first subject we can ask if it's possible to repeat the same pattern of conversion for subjects that have the same series of DICOM.
• We should save list of converted subjects

You have to make sure that each image submit as parameter to noiseAnalysis function have the same shape.

This bugs could append when we remove one slice to an odd number of slices image prior to avoid a bugs with eddy correction.

See toad/tasks/08-snr.py at line 54

Will be usefull to have the possibility to drag a freesurfer tree directly into a subject initial data

As far as I know, in functional and anatomical MRI preprocessing pipeline, the denoising step is optional and is done before motion correction and registration steps.

Should we add a switch in the pipeline to be able to do the denoising before eddy/fieldmap corrections ?

Conversion of phase maps should be converted in Float32, currently in Int16.

1. We need to evaluate if the brodmann map use into this project is optimal. Maybe there is others better solutions available.
2. Identify and label the brodmann areas into a lut file.
3. Find or define a RGB colormap convention into the lut file that will be less confusing than the actual colormap actually use.

tractography mrtrix qaSupplier is comment out.

• Add conversion of the encoding file (bvec and bval version of mrtrix)
• Add a way to keep in memory which subject have been converted

It would be helpful to create a function that compute the signal to noise ratio of an image.

When executing unf2toad without argument we've got an error message

unf2toad: error: too few arguments

It should gives help information instead of 'crashing'.

We need to add more anatomical information to post-filter the tractogram.

Don't hesitate to address fibers for this list.

• U-fibers crossing corpus callosum
• U-fibers crossing cerebellar peduncles

I put fibers that seem to be the easiest to remove.

Here is a list of files that we need to add a link to refere to in the results folder:

dwi_ -> upsampe/dwi_upsample.nii.gz
b -> ../03-eddy/dwi_EN464_subset_eddy.b
bvec -> ../03-eddy/dwi_subset_eddy.bvec
bval -> ../03-eddy/dwi_subset_eddy.bval
anat_ -> register/anat_freesurfer_resample.nii.gz
brain_mask -> register/brain_mask_resample.nii.gz
brodmann -> register/brodman_resample.nii.gz

fa_dipy, ad_dipy, rd_dipy, md_dipy, nufo_dipy, gfa_dipy, rgb -> tensordipy
fa_fsl, ad_fsl, rd_fsl, md_fsl -> tensorfsl
fa_mrtrix, ad_mrtrix, rd_mrtrix, md_mrtrix -> tensormrtrix

Any suggestions about the name of these links ?

We can use subjectname_* : (ex S01)

S01_anat.nii.gz
S01_dwi.nii.gz
S01_brain_mask.nii.gz
S01_dipy_fa.nii.gz

@arnaudbore
@mathieudesro
Find the best solution for combining two or more dwi images if the provide from the same session.

Do not supply html report if a task is ignored

Look like a custom environnement source file is needed to achieve that task.
I will probably need to adapt the grid engine for torque.
Matlab is not support. Lpca and aonlm denoising will not work. Use nlmeans or none
preprocessing white matters segmentation will not work eithers
convert see'm not install into the backend. Fixing that issue soon

INFO: Launching implementation of hardidipy task
INFO: Creating /scratch/bore_cardio/toad_data/HC_RC35_2/13-hardidipy directory
INFO: Starting tensors creation from dipy on /scratch/bore_cardio/toad_data/HC_RC35_2/05-preprocessing/dwi_HC_RC35_2_subset_eddy_unwarp_denoise_upsample.nii.gz
INFO: Start fODF computation

/usr/lib64/python2.7/site-packages/dipy/reconst/peaks.py:503: RuntimeWarning: invalid value encountered in double_scalars
return np.sqrt(numer / denom)
/usr/lib64/python2.7/site-packages/dipy/reconst/peaks.py:503: RuntimeWarning: invalid value encountered in double_scalars
return np.sqrt(numer / denom)
/usr/lib64/python2.7/site-packages/dipy/reconst/peaks.py:503: RuntimeWarning: invalid value encountered in double_scalars
return np.sqrt(numer / denom)
/usr/lib64/python2.7/site-packages/dipy/reconst/peaks.py:503: RuntimeWarning: invalid value encountered in double_scalars
return np.sqrt(numer / denom)
/usr/lib64/python2.7/site-packages/dipy/reconst/peaks.py:503: RuntimeWarning: invalid value encountered in double_scalars
return np.sqrt(numer / denom)
/usr/lib64/python2.7/site-packages/dipy/reconst/csdeconv.py:576: UserWarning: maximum number of iterations exceeded - failed to converge
warnings.warn(msg)
/opt/sge/default/spool/stark/job_scripts/18931: line 1: 5375 Bus error (core dumped) /usr/local/toad//bin/toad /scratch/bore_cardio/toad_data/HC_RC35_2 -l -p

Quand les images nécessaires à la tache fieldmap sont dans le dossier back-up, un getOrder() sur la tache retourne None. Si les images sont dans le dossier du sujet, getOrder() retourne la bonne valeur.

Cela peut poser un problème quand l'utilisateur interrompt le pipeline avant la tache fieldmap pour le relancer plus tard.

Do you understand waht I mean.

Good luck...

We should check dependency between tasks.

create a script to set environment variables for visualisation

Arnaud should specify a sessions name
Mathieu do the rest

That Readme should include disclaimer, copyright, references and instructions about how to use the files

Implement the use of the wildcard . to launch unf2toad

nibabel.nicom is weak and do not work great with anonymous data.

A good solution will be to replace nibabel.nicom by a wrapper around mrtrix dcminfo tools.