BIDS naming for the PSIR sequence


What would be the appropriate BIDS naming for the phase-sensitive inversion recovery (PSIR) sequence?


I’m fairly sure PSIR isn’t supported at the moment, but the quantitative MRI extension proposal is in a good position to add it.

BEP001 adds a suffix for inversion recovery for T1 mapping (_IRT1). Although it hasn’t been merged yet, BEP001 is fairly stable, so applying some of the rules introduced there should be alright. Assuming PSIR is the same as IRT1, but with additional phase reconstruction, then you could use the following structure for the files (adapted from an IRT1 example in the BEP):


If PSIR is different from IRT1, though, you may need to propose a new suffix. It might be worth commenting on the PR to request that suffix.

1 Like


after revisiting the answer (BIDS naming for the PSIR sequence - #2 by tsalo) to my original post, I would recommend against using -mag_IRT1 and -phase_IRT1, because the PSIR sequence typically outputs these two images:

  • magnitude reconstruction (unsigned, encoded between 0 and 4095)
  • ‘real’ reconstruction (signed, encoded between -4096 and 4095)

So, although the second scan is usually called ‘phase-sensitive’ reconstruction, it is not a phase image.

Since there is only one inversion pulse, I was tempted to get rid of inv-1, but if I understand the doc correctly, this field is mandatory when used with IRT1?

The PSIR sequence is heavily T1-weighted thought, so the suffix IRT1 seems appropriate.

I like the idea of adding acq-PSIR for transparency.

Here is a suggestion, please feel free to comment:




@tsalo how do you see adding a PSIR file collection with proper part values? Can we open a PR for this by following the guidelines we defined in the qmri appendix?

1 Like

I think the idea of part-mag and part-real in a single file collection is a bit of an edge case, but we don’t have any rules limiting file collections to part-mag/part-phase or part-real/part-imag pairs, so I think it’s fine to just have the pair @jcohenadad proposed. I have to check the math, but can’t you also recover phase and imaginary values from real and magnitude ones? So you could probably calculate all four images if you want, and we’d want to allow that in BIDS.

I’m a little surprised by the range of the ‘real’ reconstruction though. Isn’t that the same range as phase images from Siemens sequences?

EDIT: Just to clarify- are you thinking of adding a new suffix or just documenting the acq-PSIR option?

Recovering complex data from mag/phase may not be fully possible, given that the key is to deal with polarity restoration of the longitudinal magnetization.

I don’t remember any paired combination requirements for part entity, so pairs other than mag/phase real/imag would be possible.

I am leaning towards adding a new suffix instead of falling back on the acq entity. I think this would make a good use case for the flexibility of the file collections.

1 Like