SNR calculation differences in MRIQC

[NeuroimagingOL]

Dear MRIQC developers,

My question relates to the calculation of the SNR in MRIQC. When setting up our sequences I compared the SNR with MRIQC (23.1.0), back in 2023. Now I updated the MRIQC to 24.0.0 and found large differences in the calculation of the SNR, however, I can not find any description about changes in the SNR calculation in the documentation.

I attachted a screenshot. There is the same dataset calculated with both versions of MRIQC, in red with 24.0.0 and in blue with 23.1.0.

They values are not only higher but also show a different direction when comparing the blue and the yellow ones, the yellow seems to be worse than the blue ones. However, in version 23, the purple leads to better results than the red one, althought blue/red and yellow/purple used the same protocol.

Thus my question is what is the ground truth? In which results can I trust when choosing the best protocol? Is MRIQC taking care on multiband or singleband sequences in calculating the SNR? Is there a gold standard for comparing SNR in different sequences?

Any help would be appreciated.

Best regards,
Tina