Question

Question Posted 07/29/15:
Dear ADNI experts,

I have noticed that some series have a step called N3m. For instance series 93981 has the following images:
Image ID | Sequence
---------|-------------------------------------------
200795 | MP-RAGE
438650 | MT1; GradWarp; N3m
217978 | MPR; GradWarp; B1 Correction; N3; Scaled

I was wondering which processed image should be considered as the best. As the image ID is larger for the N3m image, I guess it has been added to the system later. I have also found some papers by ADNI members [1] which states that, if I'm correct, N3m is better than N3 + Scaled. Should I use N3m, then?

I have asked the question on the adni-data mailing list [2] but haven't received an answer.

Thanks in advance,
Mathieu

References
----------

[1] http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2886577/

[2] https://groups.google.com/forum/#!topic/adni-data/YXBG3mK_q7w
Response posted 08/04/15 by ADNI MRI Core:
Thank you for your question to the ADNI MRI Core.


Hi Mathieu

After consulting with the experts here at Mayo for ADNI answer is, "it depends".

The N3m processed data was provided for a subset of subjects that continued into ADNI-2. If the analysis has group wise comparisons then use the processing which is compatible across your subjects. For example, comparing "cross-over" subjects to an ADNI-1 cohort would probably suggest using the older version. If you are comparing them with ADNI-2 subjects then N3m is likely more compatible. As a rule-of-thumb, pick versions to make the preprocessing homogeneous in your analysis. With respect to spatial distortion correction, these should be indistinguishable (if not identical).

Hope this helps…


ADNI MRI
Response posted 08/04/15 by ADNI MRI Core:
Thank you for your question to the ADNI MRI Core.


Hi Mathieu

After consulting with the experts here at Mayo for ADNI answer is, "it depends".

The N3m processed data was provided for a subset of subjects that continued into ADNI-2. If the analysis has group wise comparisons then use the processing which is compatible across your subjects. For example, comparing "cross-over" subjects to an ADNI-1 cohort would probably suggest using the older version. If you are comparing them with ADNI-2 subjects then N3m is likely more compatible. As a rule-of-thumb, pick versions to make the preprocessing homogeneous in your analysis. With respect to spatial distortion correction, these should be indistinguishable (if not identical).

Hope this helps…


ADNI MRI