Differentiating Creatine and Phosphocreatine In Vivo Using 3 T 1H MR Spectroscopy

TL;DR

This paper improves the accuracy of measuring creatine and phosphocreatine in the body using magnetic resonance spectroscopy.

Contribution

The study introduces modified settings in LCModel to reduce bias in phosphocreatine quantification at 3 T.

Findings

Default LCModel settings provide accurate estimates without non-metabolite background.

Bias occurs when non-metabolite background is present, depending on lineshape and SNR.

Flexible LCModel settings mitigate bias and improve phosphocreatine quantitation.

Abstract

Validation and improvements to the proton MR spectroscopic measurement of creatine and phosphocreatine in vivo at 3 T.

Spectral simulations were used to evaluate and explore the mitigation of LCModel fitting bias due to lineshape, signal‐to‐noise (SNR) and non‐metabolite backgrounds with a focus on phosphocreatine and the spline regularization settings in the fitting model.

Although phosphocreatine and creatine are visually unresolved, excellent fitting estimates are achieved in the absence of non‐metabolite background using default settings. However, in the presence of non‐metabolite background, bias is observed and depends on lineshape, SNR and differences in the non‐metabolite background. A more flexible setting for the LCModel spline largely mitigates this bias, allowing for the confidence to use proton MRS for PCr dynamic studies.

Without prior knowledge of the non‐metabolite signal, default settings of LCModel can introduce significant bias. Modified setup for LCModel improves quantitation of phosphocreatine and reduces bias in the overall set of metabolites.

Figures