Improving spectra quality

MRS Software Osprey
1

Hi everyone,

I’m currently working with Osprey on a dataset that includes patients with ASD, and I’m using MEGA-PRESS to quantify GABA.

I’ve been using the job file provided by Osprey, but I’m running into some issues: a large number of spectra are of very poor quality.
jobDAT_MEGA.m (14.1 KB)

I’m attaching screenshots of the spectra so you can get a better idea of what I’m dealing with.
PRESS spectra.pdf (358.1 KB)
PRESS spectra 2.pdf (343.2 KB)
spectra MEGA PRESS.pdf (327.7 KB)
MEGA PRESS spectra 2.pdf (315.7 KB)

I’m also sharing some quality metrics from Osprey for context (mean, SD, min, max across subjects):

data quality
data quality1140×400 71.3 KB

A few questions:

  1. Do you have any tips or techniques to improve spectral quality ?
  2. Would it be possible / relevant to use Physics-Informed Neural Networks (PINNs) in this kind of pipeline?

Thanks a lot for your help!

@Frederic