Dear @DivyanshuTak,
Congratulations on the amazing work!
I was wondering if you could clarify a detail regarding the preprocessing pipeline.
For the specific task of overall survival prediction, was the preprocessing pipeline kept the same as the one described in the repository?
Specifically, was the pipeline:
DICOM → NIfTI → N4 bias correction (was this step persisted in the final data?)
→ resampling to 1 mm³ isotropic resolution
→ rigid registration to MNI space
→ skull stripping (HD-BET, as the final step)
From the GitHub code, it seems that N4 bias correction is indeed applied in mri_preprocess_3d_simple.py:
https://github.com/AIM-KannLab/BrainIAC/blob/main/src/preprocessing/mri_preprocess_3d_simple.py
However, my confusion comes from the downstream analysis notebook, where the input data are briefly described as:
- MRI format: NIfTI (
.nii or .nii.gz)
- Image size:
96×96×96 voxels (automatically resized)
- Sequences: T1, T1CE, T2, FLAIR (all four required)
Could you please clarify whether the downstream analysis uses the outputs of the preprocessing pipeline above?
Thank you very much for your time and for sharing this work.
Best regards,
Ines C
Dear @DivyanshuTak,
Congratulations on the amazing work!
I was wondering if you could clarify a detail regarding the preprocessing pipeline.
For the specific task of overall survival prediction, was the preprocessing pipeline kept the same as the one described in the repository?
Specifically, was the pipeline:
From the GitHub code, it seems that N4 bias correction is indeed applied in
mri_preprocess_3d_simple.py:https://github.com/AIM-KannLab/BrainIAC/blob/main/src/preprocessing/mri_preprocess_3d_simple.py
However, my confusion comes from the downstream analysis notebook, where the input data are briefly described as:
.niior.nii.gz)96×96×96voxels (automatically resized)Could you please clarify whether the downstream analysis uses the outputs of the preprocessing pipeline above?
Thank you very much for your time and for sharing this work.
Best regards,
Ines C