André Döring^{1}, Derek K Jones^{1}, and Roland Kreis^{2,3}

^{1}Cardiff University Brain Research Imaging Centre (CUBRIC), Cardiff University, Cardiff, United Kingdom, ^{2}Magnetic Resonance Methodology, Institute of Diagnostic and Interventional Neuroradiology, University Bern, Bern, Switzerland, ^{3}Translational Imaging Center, sitem-insel, Bern, Switzerland

We demonstrate that, even without having water as an internal reference, a combination of spectral registration and fitting can restore artificial signal loss promoted by incoherent averaging due to frequency/phase drifts and motion-induced dephasing for the cardinal brain metabolites (tCr, tCho, tNAA, Glx and Ins) by using a set of synthetically distorted diffusion MR spectra (including realistic phase, frequency and amplitude fluctuations).

__Amplitude Loss__mimicking effects of non-linear motion realized by rescaling individual acquisitions according to a*Weibull distribution*keeping 5/32 shots (≈15%) above a signal-level of 98% (portion of unaffected acquisitions). Signal decay gradually amplified with b-value (at highest b-value maximum decay to below 20%).__Phase Distortions__with a constant but randomly-selected phase offset between ±30° for each b-value and modulated with a*uniformly distributed*random variation of phases for single acquisitions (SD increases from 15° at b_{min}to 400° at b_{max}).__Frequency Shifts__modeled by a gradually increasing constant offset (0Hz at b_{min}to 30Hz at b_{max}) plus random*normally distributed*offsets for single acquisitions (SD 10Hz).__White Gaussian Noise__random at each acquisition and increased in amplitude to create a set of 6 noise realizations with gradually worsening SNR.

__Pre-Frequency alignment__realized by identifying and aligning NAA peak in the modulus spectrum of each acquisition, broadened by 10Hz Gauss.__Spectral registration (FID-A)__done in time-domain with*FID-A*^{6}. For b=0s/mm² single scans aligned to the best shot with the lowest unlikeness metric; for b>0s/mm² alignment referenced to the average b=0s/mm² spectrum.__Amplitude compensation (FID-A+FiTAID)__in two sub-steps:**(i)**Single shots fitted to the initial averaged spectrum from step 2 to estimate remaining frequency and phase offsets, but also amplitude fluctuations.**(ii)**These offset data were imported into MatLab for fine-tuned frequency/phase alignments and to apply signal rescaling to the 15% quantile of shots with the highest amplitude to compensate for motion^{1}. Estimated areas used as weights for weighted averaging (square of the noise)^{1}.

Comparing spectra corrected with

If motion remains uncorrected, as in the standard dMRS

Fig. 5B juxtaposes both correction approaches where the Root-Mean-Square-Relative-Errors

A limitation of this study should be noted: synthetically generated distortions do not include stronger eddy-currents and line broadening at higher b-values.

- Döring A, Adalid V, Boesch C and Kreis R. Diffusion-weighted magnetic resonance spectroscopy boosted by simultaneously acquired water reference signals. Magn Reson Med 2018, 80:2326–38.
- Najac C, Ligneul C, Branzoli F, Palombo M and Valette J. Best practices and tools for Diffusion MR Spectroscopy - Pregame Fitting Challenge 2021, https://github.com/dwmrshub/pregame-workshop-2021.
- Simpson R, Devenyi GA, Jezzard P, Hennessy TJ and Near J. Advanced processing and simulation of MRS data using the FID appliance (FID‐A) - An open source, MATLAB‐based toolkit. Magn Reson Med 2017, 77:23–33.
- Adalid V, Döring A, Kyathanahally SP, Bolliger CS, Boesch C and Kreis R. Fitting interrelated datasets: metabolite diffusion and general lineshapes. MAGMA 2017, 30:429–48.
- Şimşek K, Döring A, Pampel A, Möller HE and Kreis R. Diffusion-weighted MRS at short TE using a Connectom system: non-Gaussian metabolite diffusion and macromolecular signals in human brain. ISMRM 2020, #364.
- Near J, Edden R, Evans CJ, Paquin R, Harris A and Jezzard P. Frequency and phase drift correction of magnetic resonance spectroscopy data by spectral registration in the time domain. Magn Reson Med 2015, 73:44–50.
- Göçken M, Özçalıcı M, Boru A and Dosdoğru AT. Integrating metaheuristics and Artificial Neural Networks for improved stock price prediction. Expert Syst Appl 2016, 44:320–31.

DOI: https://doi.org/10.58530/2022/1078