Mapping QSIPrep CNR columns to b-value shells

Hi everyone,

I’m working with a multishell dMRI dataset in which each subject has three separate diffusion images per shell: b = 700, 1000, and 2000. I preprocessed the data with QSIPrep 1.1.1, which generates a combined preprocessed file with a *desc-image_qc.tsv file.

I would like to extract the median CNR for each shell as a QC metric before group analysis. However, in the QC file, the CNR columns are named only as CNR0_median, CNR1_median, CNR2_median, and CNR3_median, and I am not sure how these indices map onto the actual b-value shells.

In the preprocessed output .bval file, the nonzero shells mostly appear in the order b=1000, b=2000, and b=700, with b0 volumes interleaved. I initially thought the CNR columns might follow the shell order in the .bval file, but I am not sure.

For this specific subject, the values are approximately:

CNR0 = 16.58
CNR1 = 0.66
CNR2 = 1.81
CNR3 = 5.39

Based on the values alone, I would not want to assume the mapping, although CNR0 may correspond to b0 and the other columns to the diffusion shells. Does anyone know how QSIPrep/DSI Studio orders these CNR columns? Are they ordered by b-value, by first appearance in the .bval file, or in some other way? I suspect it is in ascending b-value order, given that it is generated by eddy.

Thanks!

Amir

Hi @AmirHussein,

The CNR[0-N] are in order of acending b-value magnitude. So in your case 0:0, 1:700, 2:1000, 3:2000.

Best,

Steven

1 Like