Dataset JSON mismatches and missing text prompts for several subsets
First, thank you for all the hard work you’ve put into preparing this challenge dataset.
While loading the files, I noticed that a couple of dataset folders don’t match the names used in the prompts JSON file. The abdominal CT folder is called CT_Abdomen1K, but the JSON refers to it as CT_AbdomenCT1K. Similarly, the autoPET folder appears in the JSON as PET_autoPET. These typos are easy enough to fix locally, but they currently prevent the loader from finding the correct prompts.
I also found that four datasets—CT_KidneyTumor, CT_TotalSeg‑Ribs, MR_HeadHaN, and MR_Spine2014—have no prompt entries in the JSON at all, which means they can’t be used for prompt‑based training or evaluation.
For now I’m excluding those subsets and patching the two typos by hand, but it would be great to get an updated JSON so everyone benefits from a consistent reference. Please let me know if you need any additional information or testing on my end.