search

StanfordMIMI / Comp2Comp

Computed tomography to body composition (Comp2Comp).
Apache License 2.0
59 stars 11 forks source link

Possible to work on google colab? #86

Open Meddebma opened 1 year ago

Meddebma commented 1 year ago

Hi,

is it possible to run comp2comp on google colab? and do you consider to integrate nifti format for input and output images?

Best, Aymen

Meddebma commented 1 year ago

Thank you so much for this code, it worked really well. I am impressed with the quality!

louisblankemeier commented 1 year ago

Hi @Meddebma,

Did you get the code running on google colab? I haven't tried this so curious if it worked for you. Thank you!

louisblankemeier commented 1 year ago

Also, we have a function that supports nifti inputs here https://github.com/StanfordMIMI/Comp2Comp/blob/master/comp2comp/io/io_utils.py#L31 but have not yet tested this extensively as a part of the pipelines. We will at this to our todos. We do not currently support nifti outputs, besides the segmentation files which are stored as niftis.

Meddebma commented 1 year ago

Hi @louisblankemeier ,

yes it worked perfectly, here is the code: !git clone https://github.com/StanfordMIMI/Comp2Comp/ %cd Comp2Comp !pip install -e . !bin/C2C spine_muscle_adipose_tissue --input_path /content/drive/MyDrive/CT_body_comp_AI/example/sr It took 90-120sec per CT scan. I am really impressed how good it is! Definitely gonna use it for my next research

louisblankemeier commented 1 year ago

I'm glad it worked for you on colab, that's really good to know!!

akshaysc commented 1 year ago

Maybe a Colab version is something we post on the repo as an example use case? Perhaps with voxel integration for viewing?

louisblankemeier commented 1 year ago

Agreed!! I think we've surfaced several cases where people would want to interact with outputs in python (going beyond the CL interface) and a colab tutorial would be well suited here. As you mentioned, one case is visualizing outputs for development and QA, and voxel would be perfect for that! I think we should add this to the top of our TODOs.