Correcting frequency and phase offsets in MRS data using robust spectral registration

Mikkelsen, Mark E.; Tapper, Sofie; Near, Jamie; Mostofsky, Stewart H.; Puts, Nicolaas A.J.; Edden, Richard A.E.

doi:10.1002/nbm.4368

Cited by 53 publications

(67 citation statements)

References 34 publications

(56 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Receiver‐coil combination 20 and eddy‐current correction 21 of the metabolite data were performed using the water reference data. Individual transients in Siemens and GE data were frequency‐ and phase‐aligned using robust spectral registration 22 . The Philips data had been coil‐combined by weighted combination with the complex coefficients obtained during the survey scan and averaged on the scanner without frequency and phase correction of the individual transients.…”

Section: Methodsmentioning

confidence: 99%

“…Individual transients in Siemens and GE data were frequency-and phasealigned using robust spectral registration. 22 The Philips data had been coil-combined by weighted combination with the complex coefficients obtained during the survey scan and averaged on the scanner without frequency and phase correction of the individual transients. After averaging F I G U R E 1 Voxel position and overview of the MRS analysis pipeline.…”

Section: Data Preprocessingmentioning

confidence: 99%

See 1 more Smart Citation

Comparison of different linear‐combination modeling algorithms for short‐TE proton spectra

et al. 2021

Self Cite

View full text Add to dashboard Cite

Short‐TE proton MRS is used to study metabolism in the human brain. Common analysis methods model the data as a linear combination of metabolite basis spectra. This large‐scale multi‐site study compares the levels of the four major metabolite complexes in short‐TE spectra estimated by three linear‐combination modeling (LCM) algorithms. 277 medial parietal lobe short‐TE PRESS spectra (TE = 35 ms) from a recent 3 T multi‐site study were preprocessed with the Osprey software. The resulting spectra were modeled with Osprey, Tarquin and LCModel, using the same three vendor‐specific basis sets (GE, Philips and Siemens) for each algorithm. Levels of total N‐acetylaspartate (tNAA), total choline (tCho), myo‐inositol (mI) and glutamate + glutamine (Glx) were quantified with respect to total creatine (tCr). Group means and coefficient of variations of metabolite estimates agreed well for tNAA and tCho across vendors and algorithms, but substantially less so for Glx and mI, with mI systematically estimated as lower by Tarquin. The cohort mean coefficient of determination for all pairs of LCM algorithms across all datasets and metabolites was trueR2¯= 0.39, indicating generally only moderate agreement of individual metabolite estimates between algorithms. There was a significant correlation between local baseline amplitude and metabolite estimates (cohort mean trueR2¯= 0.10). While mean estimates of major metabolite complexes broadly agree between linear‐combination modeling algorithms at group level, correlations between algorithms are only weak‐to‐moderate, despite standardized preprocessing, a large sample of young, healthy and cooperative subjects, and high spectral quality. These findings raise concerns about the comparability of MRS studies, which typically use one LCM software and much smaller sample sizes.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Data Preprocessingmentioning

confidence: 99%

Comparison of different linear‐combination modeling algorithms for short‐TE proton spectra

et al. 2021

Self Cite

View full text Add to dashboard Cite

show abstract

“…Four different alignment methods were performed on all data: 1) creatine (Cr) peak alignment, 2) N-acetylaspartate (NAA) peak alignment, 3) spectral registration (SR; Near et al, 2015 ), and 4) robust spectral registration (rSR; Mikkelsen et al, 2020 ). We also present results from data with no alignment performed.…”

Section: Methodsmentioning

confidence: 99%

“…Test-retest reliability has been measured ( Brix et al, 2017 ; Saleh et al, 2016 ), but does not speak to accuracy. Alignment algorithms have often been developed with reference to their success in reducing subtraction artifacts ( Evans et al, 2013 ; Mikkelsen, Saleh, et al, 2018 ; Mikkelsen et al, 2020 ), which is reasonable, but again offers no indication of accuracy. There is some evidence that the concentration of certain pairs of metabolites is related, e.g., GABA and Glx ( Steel et al, 2020 ), so it would be possible to use the strength of the relationship between two related metabolites as an index of accuracy.…”

Section: Introductionmentioning

confidence: 99%

Comparison of methods for spectral alignment and signal modelling of GABA-edited MR spectroscopy data

2021

Self Cite

View full text Add to dashboard Cite

Many methods exist for aligning and quantifying magnetic resonance spectroscopy (MRS) data to measure in vivo γ -aminobutyric acid (GABA). Research comparing the performance of these methods is scarce partly due to the lack of ground-truth measurements. The concentration of GABA is approximately two times higher in grey matter than in white matter. Here we use the proportion of grey matter within the MRS voxel as a proxy for ground-truth GABA concentration to compare the performance of four spectral alignment methods (i.e., retrospective frequency and phase drift correction) and six GABA signal modelling methods. We analyse a diverse dataset of 432 MEGA-PRESS scans targeting multiple brain regions and find that alignment to the creatine (Cr) signal produces GABA+ estimates that account for approximately twice as much of the variance in grey matter as the next best performing alignment method. Further, Cr alignment was the most robust, producing the fewest outliers. By contrast, all signal modelling methods, except for the single-Lorentzian model, performed similarly well. Our results suggest that variability in performance is primarily caused by differences in the zero-order phase estimated by each alignment method, rather than frequency, resulting from first-order phase offsets within subspectra. These results provide support for Cr alignment as the optimal method of processing MEGA-PRESS to quantify GABA. However, more broadly, they demonstrate a method of benchmarking quantification of in vivo metabolite concentration from other MRS sequences.

show abstract

“…Receivercoil combination 20 and eddy-current correction 21 of the metabolite data were performed using the water reference data. Individual transients in Siemens and GE data were frequency-and-phase aligned using robust spectral registration 22 , while Philips data had been averaged on the scanner. After averaging the individual transients, the residual water signal was removed with a Hankel singular value decomposition (HSVD) filter 23 .…”

Section: Methodsmentioning

confidence: 99%

Comparison of different linear-combination modelling algorithms for short-TE proton spectra

Zöllner

Považan

Hui

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

Keywords 21• Magnetic resonance spectroscopy 22 Highlights 25 • Means and CVs for tNAA and tCho are highly consistent across vendors and algorithms. 26• Means and CVs for mI and Glx are less consistent across vendors and algorithms. 27• Agreement between metabolite estimates from different algorithms is moderate at best. 28• Baseline estimation contributes significantly to measurement variance. 29Abstract 30 Short-TE proton magnetic resonance spectroscopy is commonly used to study metabolism in the 31 human brain. Spectral modelling is a crucial analysis step, yielding quantitative estimates of me-32 tabolite levels. Commonly used quantification methods model the data as a linear combination of 33 metabolite basis spectra, maximizing the use of prior knowledge to constrain the model solution. 34 Various linear-combination modelling (LCM) algorithms have been integrated into widely used 35 commercial and open-source analysis programs. 36This large-scale, in-vivo, multi-vendor study compares the levels of four major metabolite com-37 plexes estimated by three LCM algorithms (Osprey, LCModel, and Tarquin). 277 short-TE spec-38 tra from a recent multi-site study were pre-processed with the Osprey software. The resulting 39 spectra were modelled with Osprey, Tarquin and LCModel, using the same vendor-specific basis 40 sets. Levels of total N-acetylaspartate (tNAA), total choline (tCho), myo-inositol (mI), and gluta-41 mate+glutamine (Glx) were quantified with respect to total creatine (tCr). 42Mean spectra and models showed high agreement between all vendors and LCM algorithms. In 43 contrast, the algorithms differed notably in their baseline estimates and mI models. Group means 44 and CVs of the metabolite estimates agreed well for tNAA and tCho across vendors and algo-45 rithms. For mI and Glx, group means metabolite estimates and CVs agreed less well between 46 algorithms, with mI systematically estimated lower by Tarquin. Across all datasets, the metabo-47 lite-mean CV was 10.4% for Osprey, 12.6% for LCModel, and 14.0% for Tarquin. The grand 48 mean correlation coefficient for all pairs of LCM algorithms across all datasets and metabolites 49 was R 2 = 0.39, indicating generally moderate agreement of individual metabolite estimates be-50 tween algorithms. Stronger correlations were found for tNAA and tCho than for Glx and mI. 51 Correlations between pairs of algorithms were comparably moderate (Tarquin-vs-LCModel: R 2 52 = 0.43; Osprey-vs-LCModel: R 2 = 0.38; Osprey-vs-Tarquin: R 2 = 0.37). There was a signifi-53 cant association between local baseline power and metabolite estimates (grand mean R 2 = 0.10; 54 up to 0.62 for LCModel analysis of Glx in Siemens datasets). Metabolite estimates with stronger 55 associations to the local baseline power (mI and Glx) showed higher variance, suggesting that 56 baseline estimation has a stronger influence on those metabolites than on tNAA and tCho. 57While estimates of major metabolite complexes broadly agree between linear-combination mod-58 elling algorithms at group ...

show abstract

Correcting frequency and phase offsets in MRS data using robust spectral registration

Cited by 53 publications

References 34 publications

Comparison of different linear‐combination modeling algorithms for short‐TE proton spectra

Comparison of different linear‐combination modeling algorithms for short‐TE proton spectra

Comparison of methods for spectral alignment and signal modelling of GABA-edited MR spectroscopy data

Comparison of different linear-combination modelling algorithms for short-TE proton spectra

Contact Info

Product

Resources

About