MRIQC metrics as pass/fail criteria for T1w MRI

I want to use Qi1 and Qi2 from mriqc for automatic quality assessment of T1w MRI data. However, I’m having issues selecting appropriate thresholds for these indices.

I’m using this paper as a reference, where they set Qi1 and Qi2 thresholds to 5.06e–3 and 5.7e–2 respectively to classify between good and bad quality T1w images from ADNI.

I tried to use these values to classify a set of T1w images with evident artifacts from ADNI but the results were always far below the thresholds. I know that I’m not using the exact same subjects that those in the paper, but nonetheless they are from the same project and use the same acquisition protocols.

Does anyone know what could be blocking me to reproduce the results? Also, is there any other consensus regarding what values of Qi1 and Qi2 can be used as proper thresholds?

Thanks a lot.