CAT12 Quality metrics IQR export from XML

Dear CAT12 experts,
I am trying to export data from a large number of subjects processed with CAT12 by extracting from the XML output to a custom file. This has been fine except that I cannot reconcile the PDF report Bias and Noise “grading” with data from the xml file. I presume these have direct relationships with the ICR and NCR, but when comparing these to the linear grade assignments (from: CAT12 - Computational Anatomy Toolbox for SPM12) I don’t see any conversion factors.

Is there somewhere I could obtain more detailed information on these XML metrics, or how to find the methods for assigning (A-F) “grading” from these reports based on data in the XML?

Thank you very much!

Hi @Acinonyx. I can’t offer a real explanation, but based on looking through the MATLAB files and some fiddling I have found a few relations that seem right (this is for CAT12.8, revision 1931).

Values in the “linear rating scale” (0.5 - 10.5) can be converted to “rps” (0 - 100) by multiplying by 10 and subtracting the result from 105. For example, the xml file has a value qualityratings.IQR, which is related to the “Weighted Average (IQR)” in the pdf as

wighted_average = 105 - 10*qualityratings.IQR

Similarly, qualityratings.ICR from the xml is related to “bias” in the pdf like

bias = 105 - 10*qualityratings.ICR

Note that ICR does not factor into the “weighted average (IQR)”. For that overall value, the only values that matter are “noise” (105 - 10*qualityratings.NCR) and “resolution” (105 - 10*qualityratings.res_RMS)

The values in the xml are related according to

qualityratings.IQR = (0.5*qualityratings.NCR^8 + 0.5*qualityratings.res_RMS^8) ^ (1/8)

Values on the 0 - 100 scale are given grades of A - F (i.e., new letter grade every 10%, but anything below 50% is an F).

HTH

1 Like

Wow, yes that’s it. Thank you very much for your help!