Questions about the Oasis data

[taotianli]

Hello,

Thank you for sharing the code for this article. I have encountered a couple of questions while working with it:

1. I noticed that the file names in the pairwise_oasis folder don't match those listed in the configs/OASIS3_patients_condition.txt. For example, '30001_129_757' folder contain subject OAS30001's two MRIs, is the day to visit of these two MRIs is 129 and 757? But I did not find 129 and 757 in the txt file for OAS 30001. Which one should I consider as the correct reference?
2. I would like to confirm whether the training in this study was conducted using random 3-slice MRI data (e.g., 90:93 in the dataloader.py) rather than the entire 3D dataset.
3. BTW, what does the different conditions mean? 1 - Normal control, 2 - MCI, 3 - AD?

Thank you for your help!

[MattiaLitrico]

Hi, thanks for your comment.

1- You are right. In the dataset.py file at line 114, we determine the condition at the day of the scan by selecting the record, in the patients_condition, file that is temporally closer to the day of the scan and seeing the condition at the day.
2- We select the 3 central slice (90:93) in each MRI.
3- Yes.

[taotianli]

Thanks a lot for your kind explanation.