Our group is planning a Kaggle challenge where participants will create algorithms to impute (synthetically) missing tabular clinical data.
Inspired by the TADPOLE challenge (
https://adni.loni.usc.edu/tadpole-challenge-dataset-available/ ) , I was hoping to use the ADNI dataset. The idea would be to host the clean train/val/test csv files on LONI, such that all the participants would have to go through the proper access channels and sign the data usage agreement before download.
We thought ADNI clinical data would be ideal, as it is much closer to "real world data" compared to the toy missing data benchmarks we found in the literature.
Please let me know if this would be possible.