About Stopping criterion "cmc"

[zyyNUPU]

Hello!

I am a graduate student from Northwestern Polytechnical University, and my current research direction is Tractography.

Your work has aroused my great interest, and I am currently reproducing it. At present, there are some questions that I hope you can help answer.

I have read your code and found that "cmc" was not encountered in the stop criteria. After reading your latest paper, I found that the code for this GitHub library is TraceOracle-RL, and I am unable to access the previous version of Track to Learn. Could you please provide it to me? (Especially regarding the code that introduces anatomical priors). In addition, I am planning to learn from your work and promote the development of the field of Tractography, so I hope we can have closer communication. Can you provide me with your email? Thank you. (My email: yueyue.zhu@mail.nwpu.edu.cn ）

[AntoineTheb]

Hi @zyyNUPU . Every version of Track-To-Learn is listed in the README, with links to specific versions. If you are looking for the version of Track-to-Learn which includes CMC, you can find it under "Incorporating anatomical priors into Track-to-Learn": https://github.com/scil-vital/TrackToLearn?tab=readme-ov-file#versions

[zyyNUPU]

Hi @zyyNUPU . Every version of Track-To-Learn is listed in the README, with links to specific versions. If you are looking for the version of Track-to-Learn which includes CMC, you can find it under "Incorporating anatomical priors into Track-to-Learn": https://github.com/scil-vital/TrackToLearn?tab=readme-ov-file#versions

Thank you very much for your response. However, the link under the section titled "Incorporating anatomical priors into Track-to-Learn" does not point to the code.

Additionally, I have a question regarding data processing and hope you can assist me. Here are the steps I used for data processing with Mrtrix3, using the tournier07 spherical harmonics. If I now want to use the descoteaux07 spherical harmonics to process the data, what methods or steps should I use?

Dwi2mask DWI.mif mask.mif

dwi2response dhollander DWI.mif wm_response.txt gm_response.txt csf_response.txt

dwi2fod msmt_csd DWI.mif wm_response.txt wm_fod.mif gm_response.txt gm.mif csf_response.txt csf.mif -mask mask.mif -lmax 8,8,8 -nthreads 8

mrconvert wm_fod.mif -coord 3 0 wm_fod_norm.mif

mrthreshold -abs 0.1 wm_fod_norm.mif wm_mask.mif

mrconvert wm_mask.mif wm_mask.nii.gz

maskfilter wm_mask.mif dilate wm_mask_dilated.mif

mrcalc wm_mask_dilated.mif wm_mask.mif -sub wm_gm_interface.mif

mrconvert wm_gm_interface.mif wm_gm_interface.nii.gz

sh2peaks wm_fod.mif peaks.mif -num 5

mrconvert wm_fod.mif fod.nii.gz

mrconvert peaks.mif peaks.nii.gz

[AntoineTheb]

Thank you very much for your response. However, the link under the section titled "Incorporating anatomical priors into Track-to-Learn" does not point to the code.

As mentioned in the README, the commit id (dbae9305b4a3e9f21c3249121ef5dc5ed9faa899) points to the specific commit on the main branch which corresponds to the version of Track-to-Learn which includes the usage of Continuous Mask Criterion. Here is the link on Github: https://github.com/scil-vital/TrackToLearn/tree/dbae9305b4a3e9f21c3249121ef5dc5ed9faa899

If I now want to use the descoteaux07 spherical harmonics to process the data, what methods or steps should I use?

scilpy (https://github.com/scilus/scilpy) or dipy (https://dipy.org/index.html) may be used to convert SH from tournier07 basis to descoteaux07.

[zyyNUPU]

Thank you very much for your response. However, the link under the section titled "Incorporating anatomical priors into Track-to-Learn" does not point to the code.

As mentioned in the README, the commit id (dbae9305b4a3e9f21c3249121ef5dc5ed9faa899) points to the specific commit on the main branch which corresponds to the version of Track-to-Learn which includes the usage of Continuous Mask Criterion. Here is the link on Github: https://github.com/scil-vital/TrackToLearn/tree/dbae9305b4a3e9f21c3249121ef5dc5ed9faa899

If I now want to use the descoteaux07 spherical harmonics to process the data, what methods or steps should I use?

scilpy (https://github.com/scilus/scilpy) or dipy (https://dipy.org/index.html) may be used to convert SH from tournier07 basis to descoteaux07.

Thank you very much for your response. I am grateful for your help, and as I continued my studies over the past two weeks, new questions have arisen. I kindly request your assistance with the following issues:

Question 1: In the paper “What matters in reinforcement learning for tractography,” what specific algorithm was used for generating the GT of the TractoInferno dataset? In other words, how should the scoring_data be generated?

Question 2: In the dataset you provided, the scoring_data is missing the “scil_scoring_config.json” file, which prevents me from using tractometer. The provided files, default_config.json and gt_bundles_attributes.json, do not contain the keys shown in the image below. Could you please share the correct JSON file?