search

reginabarzilaygroup / Sybil

Deep Learning for Lung Cancer Risk Prediction using LDCT
MIT License
62 stars 38 forks source link

Nifti support #12

Closed tom1193 closed 5 months ago

tom1193 commented 1 year ago

Hi, Thank you for publishing this tool. I have code extending support for nifti files at https://github.com/tom1193/Sybil/tree/nifti-wrapper. I tested on the Ardila test set using dcm2niix to generate nifti files. Risk scores on the scan level are quite different (<0.1) b/w the two formats. Differences in aggregate AUC on the test set are not as bad (<0.01). I'm happy to submit a PR if you would like to add this feature and think it's "close enough". Also open to suggestions that could account for these differences in risk scores.

pgmikhael commented 1 year ago

Hi Thomas!

Thanks for contributing!!

Any idea why the risk scores would be different? Is there a way to achieve parity between the two inputs before passing them through model (we had done that for PNG and DICOM for instance)? If so, that would be ideal.

tom1193 commented 1 year ago

I implemented a 3d scale augmentation analogous to the 2d scale. This is causing inputs to be different, assuming parity before augmentation. There is parity before augmentations for the one scan I looked at, but I didn't explicitly check the entire dataset. Looking into it!

jsilter commented 5 months ago

I don't think this file format is widely used for LDCT. Happy to re-open if there is additional demand, but for now I'm closing this issue.