I was checking for possible “batch effects” in the MRI imaging data, and thus I was trying to access information about possible different scanner models where each person was scanned on.
When looking at the Data Dictionary, it seems that there was a field “CNO” in the MRI Waiver file for the site number, but when I download the MRI Waiver file indeed that information is not there anymore.
For the MRI Acquisition metadata, it also seems I don’t have this information. I can see again in the Data Dictionary that in the past there was information about “MRI Center Name”, but that’s not available anymore.
I thus would like to ask whether anyone knows if this information is not available at all or whether I’m missing something very obvious, thanks!
I’m pretty familiar with this, we did a protocol analysis based on scanner brand, model, version, etc.
What data set are you looking at?
We were not able to find any direct downloads from PPMI that had protocol data. Our process was to download the metadata CSV’s as well as the actual Imaging Data (DICOM), and extract metadata with dcm2niix:
$ dcm2niix -z y -m n -o ./foo -f “%f” ~/Downloads/PPMI/<PATH_TO_DICOM_STUDY>/
Chris Rorden’s dcm2niiX version v1.0.20241208 (JP2:OpenJPEG) Clang15.0.0 x86-64 (64-bit MacOS)
Found 80 DICOM file(s)
Warning: Discrepancy between reported (0.465s) and estimated (109.74s) repetition time (issue 560).
Warning: Weird CSA ‘ProtocolSliceNumber’ (System/Miscellaneous/ImageNumbering reversed): VALIDATE SLICETIMING AND BVECS
Convert 80 DICOM as ./foo/STUDY_ID (448x448x16x5)
Conversion required 2.430005 seconds (2.253337 for core code).
You could then use jq to cherry pick fields from every JSON file. That’s the fastest code path I can recommend for this.
I have an R script that organizes all of this from a PPMI download, but I haven’t made it open-source yet. We are looking to understand what starter projects people need, so input here would be helpful.
I’ll also add that the site numbers seem scrubbed from all of the DICOM data, as we were also unable to find Site ID’s. But if you are just looking for scanner model/version etc, you could do that with the data set. You will just not be able to identify specific machines.
I’m not sure where my code would start vs where the PPMI downloader tools (LivingPark?) stop… I just checked out LivingPark and it seems to create Nifti files? At least it has some functions to get NIFTI files… it also seems to have some function to pull MRI metadata, but it doesn’t look like that pulls from the NIFTI files…
@tiago.azevedo Please note, the CNO field is not available for download. There have been changes over the years to the PPMI Data Dictionary, including some significant changes in the past couple of years. You mentioned you had reviewed the Data Dictionary. As a general tip, if you see a possible change, I would recommend that you check the “Annotated” copy of the data dictionary (available after login under Study Docs”) which has more detail than the basic data dictionary on historic changes, including deprecated variables and mapping to prior versions of the database. Hope that helps!
@apinstein The data I’m looking at at the moment are the already preprocessed APARC/ASEG thicknesses/volumes values. I was avoiding downloading all the DICOM files given how big they are, and that’s why I was asking whether that information would be already extracted somewhere in the LONI dashboard. I just wanted to check whether there would be any strong batch effects from the scanner brands/models/…, and I was only able to find the PROTOCOL variable as resembling this in the MRI_Acquisition_Metadatafile. I appreciate your explanations and I think I’ll eventually just download it somewhere then, thanks!
@lkirsch Thanks a lot for the tip, I did indeed checked the “Annotated” copy, and that’s why I knew the variable was not available anymore. I was just mentioning that to show what steps I took before asking the question here.