I’ve run into some issues with OspreyProcess on unedited, short TE Philips data (brainstem placement) – despite reasonable-looking metab spectra coming in, some of the processed spectra are poorly phased:
At first I suspected that the moderate frequency shift was a factor, but looking a little more closely, it apears that the ECC step is ruining the phasing:
Is there a way to unambiguously determine whether ECC has been performed on the scanner already (eg, some parameter in the .SPAR file)? The data were not collected locally, so it’s not trivial to just check at the scanner.
If it’s not recorded in the .SPAR file, would it be reasonable/feasible to infer this from the data somehow (eg: heuristic approach which concluces that ECC is unnecessary if it makes things worse)?
I notice that OspreyProcess does a final phasing to Cho/Cr, but only for unedited Siemens data (and MRSI). What’s the rationale behind this – is there any reason not to do this for other vendors’ datasets? Adding this step does improve processing outcomes in our scenario, albeit at risk of masking the underlying issue…
Great to hear from you. If water reference data are acquired for non-edited scans (i.e., if Spectral Correction is active), then the ECC is always already applied on the scanner. The Hopkins editing patches do not apply ECC on the scanner, which is why we have the option to set this flag in the job file. You should set this flag to 0. As you can see, the ‘pre-Klose’ scans look pretty good.
This information is unfortunately not recorded in the SPAR files.
The additional Siemens phasing step is, I believe, a remnant of some phasing issues that we saw when we first dragged the 275 PRESS datasets from “Big GABA/Big PRESS” through. Older Siemens sequences do not automatically acquire a water reference, so you need to acquire a separate scan, and if this separate water reference isn’t perfectly matched to the water-suppressed acquisition (for example, acquired with the water suppression module completely removed as opposed to just setting ‘rf off’ but leaving the gradients on), the eddy-current correction will, in fact, be incorrect and introducing phase errors that need to be pulled straight. There is less freedom to screw the water reference acquisition up on Philips, and unless you force it otherwise, the water reference will be acquired correctly.
