MIC-DKFZ
Longitudinal option / ideas for nnUNet?
Hello, I have used nnUNet a lot, it's a great package. I have data which is longitudinal (i.e., the same subject at two different time points). This is a fairly common use case for imaging. The question I have, is there a way to segment both images simultaneously with nnUNet? In essence, condition the segmentation of time point 2 on time point 1? I have seen other groups do this for other packages (namely brain imaging), where they claim one is able to obtain more consistent segmentations this way. Do you have any ideas on how to do that with nnUNet (beyond obviously just sequentially running it on the two time points), or any plans to incorporate this? Thanks again for your help!
Hey @karllandheer, currently nnU-Net does not support longitudinal images, but you @mrokuss and @ykirchhoff may be able to help you out in how to use longitudinal data
Hey @karllandheer
We've been working on a longitudinal version of nnUNet which will hopefully be released soonish in order to exploit exactly the benefits of time series data you were describing - will keep you posted :)
@karllandheer This is the associated repo. Feel free to open an issue there to get @mrokuss and @ykirchhoff to fill it 😉 https://github.com/MIC-DKFZ/Longitudinal-Difference-Weighting
Hey @karllandheer
If you're still interested in longitudinal segmentation, we just released LongiSeg—a framework specifically designed for longitudinal image segmentation based on nnUNet! 🚀
LongiSeg enables structured processing of temporal medical images, allowing you to incorporate multiple 3D volumes per patient. It also gives you the option to leverage our Difference Weighting Block to enhance learning from temporal changes.
You can check it out here:
🔗 LongiSeg: GitHub
🔗 Difference Weighting Block: GitHub
We hope this helps! Let us know if you have any questions or feedback. 😊
Hello, I am still interested and thanks so much for your great package and continued support of it! I will check it out!