Missing period of FDG images; disagreement of scan dates and EXAMDATE

[Tom Earnest]

Dear ADNI Google Group:

 

Our laboratory is pursuing machine learning of imaging for the representation and prediction of dementia.   ADNI is our primary source for training data for preliminary work which seeks features of increasing CDR in repeated measurements by FDG PET.  The attached image illustrates for FDG PET the relationship of the EXAMDATE variable, drawn from ADNIMERGE.csv and REGISTRY.csv, with the acquisition date provided by LONI IDA search (https://ida.loni.usc.edu/pages/access/search.jsp) in search-specific csv-tables.  

EXAMDATE is highly correlated with acquisition date through 24 Apr 2014, is unlisted in ADNIMERGE and REGISTRY through 12 Nov 2015, and following 12 Nov 2015, is discrepant from the acquisition date by as much as 37 months.  Since Jan 2019, EXAMDATE appears better correlated with acquisition date.  We are aware that errors have been noted in ADNIMERGE.EXAMDATE as described in, e.g., https://groups.google.com/g/adni-data/search?q=EXAMDATE.  The recommendation is to refer to REGISTRY.EXAMDATE for corrections or missing data.  However, the figure above arises from the union of ADNIMERGE.EXAMDATE and REGISTRY.EXAMDATE with REGISTRY values always chosen over ADNIMERGE.  

We regard the acquisition date to be the most reliable source for imaging date-times.  In spot-checks, acquisition date-times were identical to date-timed session folders in downloads (e.g., “ADNI/128_S_0230/Coreg,_Avg,_Std_Img_and_Vox_Siz,_Uniform_Resolution/2006-04-11_12_28_25.0/I57592/*.dcm”).  Acquisition date-times were also identical to DICOM fields AcquisitionDate and AcquisitionTime found in downloads.  

EXAMDATE has been described in the ADNI Google Group as the preferred temporal metric for cross-sectional and longitudinal studies, but we are concerned that our use of EXAMDATE does not appear usable for FDG PET.  We would appreciate help identifying well-known artifacts or errors in our use of EXAMDATE.  We would also value any corroboration of our findings, in FDG PET or other ADNI data with date-times.

Thank you,

John Lee, Tom Earnest

Washington University in Saint Louis

[Danielle J Harvey]

Hi Tom and John,

 

I would say that you should use the date that makes the most sense for your outcome. If FDG PET is the outcome, then scandate makes the most sense. For cognitive/clinical outcomes, the date of the visit (EXAMDATE from REGISTRY.csv) likely makes the most sense. Timing of the scan acquisition does vary from the clinic date due to scheduling challenges, etc.

 

Danielle

--
You received this message because you are subscribed to the Google Groups "Alzheimer's Disease Neuroimaging Initiative (ADNI) Data" group.
To unsubscribe from this group and stop receiving emails from it, send an email to adni-data+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/adni-data/76b10c9c-dc0d-4e50-b6f8-666a23917656n%40googlegroups.com.

**CONFIDENTIALITY NOTICE** This e-mail communication and any attachments are for the sole use of the intended recipient and may contain information that is confidential and privileged under state and federal privacy laws. If you received this e-mail in error, be aware that any unauthorized use, disclosure, copying, or distribution is strictly prohibited. If you received this e-mail in error, please contact the sender immediately and destroy/delete all copies of this message.

[Tom Earnest]

Danielle,

Thanks for the quick response!  We were struggling with using the visit codes to relate scan data to contemporaneous clinical/demographic variables.  But I think we have settled on just comparing the dates of the scans against the dates of other data collection (regardless of what visit things were assigned to).  And it sounds like what you are saying agrees with that.

Though I am still curious about the gap in FDG scans from approximately 2014-2017 (shown in the scatterplot I previously attached).  We haven't seen this same pattern for other imaging modalities, and haven't found any related documentation from ADNI.   Do you know anything about what might be going on here?

Best, 

Tom

[Danielle J Harvey]

Hi Tom,

 

Yes, probably the best is to link the scandate with the closest clinical visit to get other information for your models.

 

There are a few things that may be going on with FDG. First, FDG was being collected every 2 years starting in ADNI-GO/2. At some point, during ADNI-2 (and I apologize that I don’t recall the exact date), FDG was stopped due to budgetary constraints. In ADNI-3, I believe the plan was to collect one FDG scan. I’m not sure if any of those explain what you are seeing.

To view this discussion on the web visit https://groups.google.com/d/msgid/adni-data/e4c92c1a-a6ff-4800-bead-38d0a04a22d2n%40googlegroups.com.

[Tom Earnest]

Danielle,

Ah, sounds like the stoppage of FDG due to budget is what we are seeing!  From the image search, I am seeing that FDG scans stop around June 2014 (in the midst of ADNI2) and then resume in January 2017 (after ADNI3 started).  So that seems pretty consistent with what you described.  Thank you, we were not aware of that.  

Best,

Tom