Results and interpretation of a fitting challenge for MR spectroscopy set up by the MRS study group of ISMRM

Marjańska, Małgorzata; Deelchand, Dinesh K.; Kreis, Roland

doi:10.1002/mrm.28942

Cited by 38 publications

(53 citation statements)

References 32 publications

(103 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, low variance or AICs do not reflect higher accuracy and may rather indicate insufficient responsiveness to real biological variance. The use of simulated spectra as a ground truth is extremely promising as a validation strategy 47 , but can only succeed if synthetic data are truly representative of in-vivo conditions. Another limitation is the single LCM algorithm that we used here, Osprey, which—while conceptually and in performance 6 similar to LCModel does not feature regularization terms on the baseline or the lineshape, and features fewer metabolite soft constraints.…”

Section: Discussionmentioning

confidence: 99%

Cohort-mean measured macromolecules lead to more robust linear-combination modeling than parameterized and subject-specific ones

Zöllner

Davies-Jenkins

Murali‐Manohar

et al. 2022

Preprint

View full text Add to dashboard Cite

Expert consensus recommends linear-combination modeling (LCM) of proton MR spectra with sequence-specific simulated metabolite basis function and experimentally derived macromolecular (MM) basis functions. Measured MM basis functions have been derived from metabolite-nulled spectra averaged across a small cohort. The use of subject-specific instead of cohort-averaged measured MM basis functions has not been studied. Furthermore, measured MM basis functions are not widely available to non-expert users, who commonly rely on parameterized MM signals internally simulated by LCM software. To investigate the impact of the choice of MM modeling, this study, therefore, compares metabolite level estimates between different MM modeling strategies (cohort-mean measured; subject-specific measured; parameterized) in a lifespan cohort and characterizes its impact on metabolite-age associations. 102 conventional (TE = 30 ms) and metabolite-nulled (TI = 650 ms) PRESS datasets, acquired from the medial parietal lobe in a lifespan cohort (20-70 years of age), were analyzed in Osprey. Short-TE spectra were modeled in Osprey using six different strategies to consider the macromolecular baseline. Fully tissue- and relaxation-corrected metabolite levels were compared between MM strategies. Model performance was evaluated by model residuals, the Akaike information criterion (AIC), and the impact on metabolite-age associations. The choice of MM strategy had a significant impact on metabolite level estimates. The estimates agreed relatively well between different MM strategies (r > 0.6) and had no major impact on the variance. The lowest relative model residuals and AIC values were found for the cohort-mean measured MM. Metabolite-age associations were consistently found for two major singlet signals (tCr, tCho) for all MM strategies, but varied substantially for highly J-coupled metabolites. A variance partition analysis indicated that up to 44% of the total variance was related to the choice of MM strategy. Additionally, the variance partition analysis reproduced the metabolite-age association for tCr and tCho found in the simpler correlation analysis. In summary, the inclusion of a single high-SNR MM basis function (cohort-mean) leads to more robust metabolite estimation (lower model residuals and AIC values) compared to MM strategies with more degrees of freedom (Gaussian parametrization) or subject-specific MM information. Integration of multiple LCM analyses into a single statistical model potentially improves the robustness in the detection of underlying effects (e.g. metabolite vs age), reduces algorithm-based bias, and estimates algorithm-related variance.

show abstract

Section: Discussionmentioning

confidence: 99%

Cohort-mean measured macromolecules lead to more robust linear-combination modeling than parameterized and subject-specific ones

Zöllner

Davies-Jenkins

Murali‐Manohar

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…There are some further caveats with respect to LCModel fitting (likely also common to other similar spectral fitting software). The contributions from baseline and macromolecules present a potential source of bias, along with linewidth and SNR (see Marja nska et al 43 ). For this report, estimates of baseline and macromolecules are limited to the default LCModel software (version 6.3-1 J).…”

Section: Discussionmentioning

confidence: 99%

Proton magnetic resonance spectroscopy assessment of neonatal brain metabolism during cardiopulmonary bypass surgery

Spielman

Hurd

et al. 2022

NMR in Biomedicine

View full text Add to dashboard Cite

Here, we report on the development and performance of a robust 3‐T single‐voxel proton magnetic resonance spectroscopy (1H MRS) experimental protocol and data analysis pipeline for quantifying brain metabolism during cardiopulmonary bypass (CPB) surgery in a neonatal porcine model, with the overall goal of elucidating primary mechanisms of brain injury associated with these procedures. The specific aims were to assess which metabolic processes can be reliably interrogated by 1H MRS on a 3‐T clinical scanner and to provide an initial assessment of brain metabolism during deep hypothermia cardiac arrest (DHCA) surgery and recovery. Fourteen neonatal pigs underwent CPB surgery while placed in a 3‐T MRI scanner for 18, 28, and 37°C DHCA studies under hyperglycemic, euglycemic, and hypoglycemic conditions. Total imaging times, including baseline measurements, circulatory arrest (CA), and recovery averaged 3 h/animal, during which 30–40 single‐voxel 1H MRS spectra (sLASER pulse sequence, TR/TE = 2000/30 ms, 64 or 128 averages) were acquired from a 2.2‐cc right midbrain voxel. 1H MRS at 3 T was able to reliably quantify (1) anaerobic metabolism via depletion of brain glucose and the associated build‐up of lactate during CA, (2) phosphocreatine (PCr) to creatine (Cr) conversion during CA and subsequent recovery upon reperfusion, (3) a robust increase in the glutamine‐to‐glutamate (Gln/Glu) ratio during the post‐CA recovery period, and (4) a broadening of the water peak during CA. In vivo 1H MRS at 3 T can reliably quantify subtle metabolic brain changes previously deemed challenging to interrogate, including brain glucose concentrations even under hypoglycemic conditions, ATP usage via the conversion of PCr to Cr, and differential changes in Glu and Gln. Observed metabolic changes during CPB surgery of a neonatal porcine model provide new insights into possible mechanisms for prevention of neuronal injury.

show abstract

“…Our simulated dataset, containing 96000 signals, was generated using Eq. 1 and the publicly available metabolite basis set from the ISMRM MRS study group’s fitting challenge (Marjańska et al, 2022) (19 metabolites signals (see Table 1) and one MM signal, 3T, PRESS, TE = 30 ms, spectral width = 4000 Hz, 2048 points). The range of parameters ( A m , Δ α , Δ f , and Δ θ ) were determined according to the literature (Govindaraju et al, 2000; Robin A. de Graaf, 2019) (listed in Table 1).…”

Section: Methodsmentioning

confidence: 99%

Physics-informed Deep Learning Approach to Quantification of Human Brain Metabolites from Magnetic Resonance Spectroscopy Data

Shamaei

Starčuková

Starčuk

2022

Preprint

View full text Add to dashboard Cite

Purpose: While the recommended analysis method for magnetic resonance spectroscopy data is linear combination model (LCM) fitting, the supervised deep learning (DL) approach for quantification of MR spectroscopy (MRS) data recently showed encouraging results; however, supervised learning requires ground truth fitted spectra, which is not practical. This work investigates the feasibility and efficiency of the LCM-based self-supervised DL method for the analysis of MRS data. Method: We present a novel DL-based method for the quantification of relative metabolite concentrations, using quantum-mechanics simulated metabolite responses and neural networks. We trained, validated, and evaluated the proposed networks with simulated and publicly accessible in-vivo human brain MRS data and compared the performance with traditional methods. A novel adaptive macromolecule fitting algorithm is included. We investigated the performance of the proposed methods in a Monte Carlo (MC) study. Result: The validation using low-SNR simulated data demonstrated that the proposed methods could perform quantification comparably to other methods. The applicability of the proposed method for the quantification of in-vivo MRS data was demonstrated. Our proposed networks have the potential to reduce computation time significantly. Conclusion: The proposed model-constrained deep neural networks trained in a self-supervised manner with complex data can offer fast and efficient quantification of MRS data.

show abstract

Results and interpretation of a fitting challenge for MR spectroscopy set up by the MRS study group of ISMRM

Cited by 38 publications

References 32 publications

Cohort-mean measured macromolecules lead to more robust linear-combination modeling than parameterized and subject-specific ones

Cohort-mean measured macromolecules lead to more robust linear-combination modeling than parameterized and subject-specific ones

Proton magnetic resonance spectroscopy assessment of neonatal brain metabolism during cardiopulmonary bypass surgery

Physics-informed Deep Learning Approach to Quantification of Human Brain Metabolites from Magnetic Resonance Spectroscopy Data

Contact Info

Product

Resources

About