Inconsistent json file from BaiduYun and Huggingface

[cheliu-computation]

Thanks for your impressive work and the effort on the dataset construction!

I have download the dataset from BaiduYun and the json file from huggingface.
The json from baidu shows 2,893 cases for Test, 32,891 cases for train. But the last case from train set has some error about unfinished strings.
However, the json from huggingface has 7,772 cases for test, 31,086 for train.
The total cases from the paper is 39,026, but the number from BaiduYun is '35,784' and '38,858' from huggingface.
This error makes reproducing work quite hardly.

So, if we want to reproduce this work based on 20 zip file from BaiduYun, which split file should we use?
Also, I found the subfolder1.zip and subfolder2.zip have some corrupted issue, even I fixed it with 'zip -F' command, I am not sure all files are restored. If authors could fix this issue, it will be really grateful.

After unzipping, it has multiple folders but not match the path from the json file.
I guess the number from json: ''/processed_images/2534/1/CT_0'', the '2534' is the correct path for the image?

The format from two different sources are also different:
the json from huggingface: "image_path": [ "/remote-home/share/data200/172.16.11.200/zhengqiaoyu//processed_file/npys/32940/1/MRI_0.nii.gz", "/remote-home/share/data200/172.16.11.200/zhengqiaoyu//processed_file/npys/32940/1/MRI_1.nii.gz",
the json from baidu: "image_path": [ "/processed_images/1/1/CT_0", "/processed_images/1/1/CT_1", "/processed_images/1/1/CT_2", "/processed_images/1/1/CT_3" ]

Seems the huggingface json point to the original medical image, but the baidu json is for jpg only?

Looking forward to your response!

[cheliu-computation]

Additionally, in the huggingface json, the file provide 'icd10s' code, but the json from baiduyun does not have this term

[qiaoyu-zheng]

I'm sorry the json file from baiduyun has some error, please follow the json file from huggingface. Because of the uploading problem, there may be some subtle matching issue in the total number, we will check it again later. However, the exisiting version is already enough for reproducing. If you meet the matching error in the dataloader, you can just simply remove these cases, this will not lead to a significant effect on the reproduction results. Again, we will fix this error later

[SZUHvern]

Hello,

Thank you very much for your work！
I wanted to check if the relevant data and JSON corrections are now available.
I eagerly waiting for your response. Many thanks!