About training details

[xychen2022]

Hello,

I have a few questions that need your clarification:

1. I installed the latest totalsegmentator and tested it on some MR images. I found the predictions are better than I expected. So, I am wondering if you included any MRI images (and other modalities except CT and MRI) into training?
2. What postprocessing steps are applied to the model predictions? Could you point me to the corresponding code?
3. Any plan to release the images and annotations that were used to train totalsegmentator v2?

Thanks!

[wasserth]

Hello, the training data does not contain any MRI images, only CT images. We are currently working on building another TotalSegmentator for MRI. But this will take some more time. There is not really any postprocessing done. We do not have plans to release the additional data which we added to train totalsegmentator v2.