Sodium quantification in skeletal muscle: comparison between Cartesian gradient-echo and radial ultra-short echo time 23Na MRI techniques

Abstract

Background: Clinical magnetic resonance imaging (MRI) studies often use Cartesian gradient-echo (GRE) sequences with ~2-ms echo times (TEs) to monitor apparent total sodium concentration (aTSC). We compared Cartesian GRE and ultra-short echo time three-dimensional (3D) radial-readout sequences for measuring skeletal muscle aTSC.

Methods: We retrospectively evaluated 211 datasets from 112 volunteers aged 62.3 ± 12.1 years (mean ± standard deviation), acquired at 3 T from the lower leg. For ²³Na MRI acquisitions, we used a two-dimensional Cartesian GRE sequence and a density-adapted 3D radial readout sequence with cuboid field-of-view (DA-3D-RAD-C). We calibrated the ²³Na MR signal using reference tubes either with or without agarose and subsequently performed a relaxation correction. Additionally, we employed a six-echo ¹H GRE sequence and a multi-echo spin-echo sequence to calculate proton density fat fraction (PDFF) and water T2. Paired Wilcoxon signed-rank test, Cohen d_z for paired samples, and Spearman correlation were used.

Results: Relaxation correction effectively reduced the differences in muscle aTSC between the two acquisition and calibration methods (DA-3D-RAD-C using NaCl/agarose references: 20.05 versus 19.14 mM; d_z = 0.395; Cartesian GRE using NaCl/agarose references: 19.50 versus 18.82 mM; d_z = 0.427). Both aTSC of the DA-3D-RAD-C and Cartesian GRE acquisitions showed a small but significant correlation with PDFF as well as with water T2.

Conclusions: Different ²³Na MRI acquisition and calibration approaches affect aTSC values. Applying relaxation correction is advised to minimize the impact of sequence parameters on quantification, and considering additional fat correction is advisable for patients with increased fat fractions.

Relevance statement: This study highlights relaxation correction's role in improving sodium MRI accuracy, paving the way for better disease assessment and comparability of measured sodium signal in patients.

Key points: • Differences in MRI acquisition methods hamper the comparability of sodium MRI measurements. • Measured sodium values depend on used MRI sequences and calibration method. • Relaxation correction during postprocessing mitigates these discrepancies. • Thus, relaxation correction enhances accuracy of sodium MRI, aiding its clinical use.

Keywords: Calibration; Magnetic resonance imaging; Muscle (skeletal); Sodium; Volunteers.

Fig. 1

a Exemplary apparent tissue sodium concentration (aTSC) maps of the lower leg acquired with a Cartesian gradient-echo (GRE) and density-adapted 3D radial readout sequence with cuboid field-of-view (DA-3D-RAD-C) sequence (acquisition parameters presented in Table 1). Note that these are uncorrected aTSC maps using the agar references for calibration. DA-3D-RAD-C ²³Na images were reconstructed using a Hamming filter. Four different reference tubes were scanned together with the lower leg and used for signal calibration. Note the differences in the signal intensities between the reference tubes with and without agarose (Ag). b Exemplary calibration curves for the Cartesian GRE and DA-3D-RAD-C images using either the reference phantoms without (blue) or with (orange) agarose. Before correcting relaxation weighting effects, there is a noticeable difference in the calibration curves based on NaCl and agarose phantoms, which can be reduced by applying a relaxation correction (dashed lines)

Fig. 2

Comparison of apparent tissue sodium concentration (aTSC) values measured in lower leg muscles (n = 1,491) using the Cartesian gradient-echo (GRE) and DA-3D-RAD-C sequences and calibrating the ²³Na MR signal either using NaCl and agarose (Ag) reference phantoms. a Without performing a relaxation correction, aTSC values differed between the two sequences and reference types. The difference between agarose and NaCl calibration was smaller for the DA-3D-RAD-C (Cohen d_z = 0.686) than for the Cartesian GRE acquisition (Cohen d_z = 0.910). b The correction of relaxation biases reduced the difference in measured aTSC values for the DA-3D-RAD-C (Cohen d_z = 0.395) and for the Cartesian GRE acquisition (Cohen d_z = 0.427). c, d aTSC values in TA muscles (n = 213) were significantly decreased compared to all other lower leg muscle regions for both sequence types (p < 0.001). Furthermore, aTSC values of the Cartesian GRE were reduced in the PER muscle compared to the GM and EDL but not for the DA-3D-RAD-C sequence. DA-3D-RAD-C, Density-adapted 3D radial readout sequence with cuboid field-of-view; EDL, Extensor digitorum longus; GM, Gastrocnemius medialis; GL, Gastrocnemius lateralis; PER, Peroneus; SOL, Soleus; TA, Tibialis anterior; TP, Tibialis posterior

Fig. 3

Correlation between apparent tissue sodium concentration (aTSC) values measured in lower leg muscles (n = 1,491) using the density-adapted 3D radial readout sequence with cuboid field-of-view (DA-3D-RAD-C) (a, b) and Cartesian gradient-echo (GRE) (c, d) sequence by calibrating the signal intensity either using the agarose (Ag) or NaCl reference phantoms. The dashed lines represent perfect correlation. If no correction for relaxation effects was performed (a, c), the measured aTSC values differed between the two reference phantom types (a Spearman rho = 0.995; c Spearman rho = 0.997). After relaxation correction (b, d), data was aligned with the dashed line (b Spearman rho = 0.995; d Spearman rho = 0.997)

Fig. 4

Correlation of apparent tissue sodium concentration (aTSC) values in lower leg muscles (n = 1,491) measured using the density-adapted 3D radial readout with cuboid field-of-view (DA-3D-RAD-C) sequence and the Cartesian gradient-echo (GRE) sequence by calibrating the signal intensity either using the agarose (Ag) (a, b) or NaCl reference phantoms (c, d). The dashed lines represent perfect correlation. Strong dependencies were found between each correlation with Spearman rho coefficients for panels (a) 0.955, (b) 0.954, (c) 0.959, and (d) 0.959

Fig. 5

Correlations between apparent tissue sodium concentration (aTSC) (including relaxation correction, normalized on agarose phantoms) and water T2 relaxation times (a, b) and proton-density fat fraction (PDFF) (c, d) measured by ¹H MRI. Note the positive, but not perfect correlation between aTSC and water T2:_: DA-3D-RAD-C, Spearman rho = 0.472, p < 0.001; Cartesian GRE, Spearman rho = 0.467, p < .001). The PDFF is less dependent on the aTSC; DA-3D-RAD-C, Spearman rho = 0.357, p < 0.001; Cartesian GRE, Spearman rho = 0.359, p < 0.001. DA-3D-RAD-C Density-adapted 3D radial readout with cuboid field-of-view, GRE Gradient-echo