Bad fitting MPFC?

Any idea what went wrong with the fitting of this spectrum from the medial prefrontal cortex? The NAA peaks are visible in the raw spectra, and have a reasonable quantification that aligns with other values from the data set, but has been nearly completely eliminated in the “aligned and averaged” and “LC model” windows. Is the spectral quality to blame? Would I have to exclude this as a failed scan? Or is this more consistent with a flaw in the fitting algorithm or something else specific to HERCULES or Osprey?

MPFC_bad fitting.pdf (466.2 KB)

Hi @Koen_Cuypers,

could you also upload the diff1, diff2, and sum spectrum from the process window? I suspect that the registration of the 4 sub-spectra is not working as expected.

Helge

All processed spectra.pdf (1.2 MB)

Hi @Helge,
Please find the spectra in the pdf.

Best,
Koen

Hi @Koen_Cuypers,

Ok so it is definitely an issue during the subspectra alignment. Have you tried the different options?

opts.SubSpecAlignment.mets = 'L2Norm';          % OPTIONS:    - 'L2Norm' (default)
                                                %             - 'L1Norm'
                                                %             - 'none'

or using

opts.SubSpecAlignment.mets = 'L2Norm'; 
opts.UnstableWater = 1;

Best,
Helge

Dear @Helge ,

Please find the (low quality pdf) file with our results attached.
Option 3 (see below) gave the best results. I guess we should now implement these lines also for analysis of the other data which did not show the problem to be consistent?

opts.SubSpecAlignment.mets = ‘L1Norm’;
opts.UnstableWater = 1;

All 20processed spectra-compressed.pdf (661.3 KB)

Hi @Koen_Cuypers,

Agreed that options 3 gives the best results. Rerunning it for all subjects makes sense for consistency.

Best,
Helge

Thanks Helge!
Best,
Koen

1 Like