Re: Missing MRI QC values, segmentation data and MRI scans

[Neil Oxtoby]

Hi Peter,

We don't know all the ins and outs of ADNI's MRI QC procedures, so I would suggest that this is an ADNI Data question rather than a TADPOLE question.

Re: FSX vs FSL, it could be that visits m06 and m24 for RID==3 failed the longitudinal processing pipeline of FreeSurfer, which is more complicated than the cross-sectional pipeline. And FreeSurfer values are known to differ between FSL and FSX, so I'm not surprised that you found differing values.

On Monday, 16 August 2021 at 13:26:08 UTC+1 Peter H wrote:

Dear Experts,

I know this is more of a ADNI-specific question, but since these can be found in the 'TADPOLE_D1_D2.csv', I wanted to ask here first.


I found some patients who do not have MRI QC data, but have segmentation data available (the columns from 'DJ' to 'DR' in the Excel sheet are empty). For example - PTID: 126_S_2405
In the official 'ADNI Procedures Manual' PDFs it says that every MRI scan goes under a QC procedure at Mayo Clinic, so that's why I thought that each MRI image should have QC values as well. Why could be that some patients simply don't have QC data available in the table?

Another thing I found is some patients only have cross-sectional QC + segmentation data, but no info regarding longitudinal scans (UCFFSL columns are empty, UCFSSX columns are filled).
In case of PTID: 011_S_0003 there are 4 entries: bl, m06, m12, m24. According to the ADNI-1 Procedures Manual, patients with AD should have MRI scans at bl, months 6, 12 and 24, but in case of 011_S_0003 the m06 and m24 longitudinal (UCSFFSL) data are missing, while there is cross-sectional (UCSFFSX) data provided for all 4 occasions.
I thought that maybe it should be the other way around: longitudinal data complete, and some deficit in case of cross-sectional data. What could be the explanation for this?

Also what I don't know is how to interpret the cross-sectional and longitudinal measurements. I thought that the longitudinal and cross-sectional QC + segmentation data for the 'bl' scan should be identical, but I found some discrepancies regarding this with many if not with all of the patients. For example at PTID: 100_S_0006 column 'ST101SV_UCSFFSL_02_01_16_UCSFFSL51ALL_08_01_16' has a value of 1264 and for the same scan 'ST101SV_UCSFFSX_11_02_15_UCSFFSX51_08_01_16' there is a 1245 value.
Why is that? I was thinking about different FreeSurfer versions giving different results.


Thank you very much for your help!

Kind regards,
Peter

[Peter H]

Hi Neil,

thank you for your help!!

Cheers,

Peter