Based on its sensitivity to cell viability and tissue microstructure, sodium MRI has the potential to provide quantitative biomarkers of tumor grade and response to therapy. For biomarkers such as total sodium concentration (TSC) to have clinical utility, they require technical and biological validation. This work focuses on the technical validation of TSC, specifically assessing the repeatability and reproducibility of measurements in healthy volunteers. The results presented here show consistent TSC measurements across repeated scans at different sites, and good repeatability and reproducibility (coefficients of variation, CoVs < 10%).
Eleven healthy volunteers (mean ± standard deviation (SD) = 32 ± 6 years, 3 females) were scanned at two sites, A and B. Six volunteers were scanned twice at both sites, and 5 were scanned twice at one site and once at the other. Repeatability (intra-site variation) and reproducibility (inter-site variation) were assessed with scans <1 hour and between 1 and 41 days apart, respectively. Sites had different scanners (GE 3 T MR750 and Philips 3 T Achieva), but both used the same model of 1H/23Na dual-tuned head coil (RAPID Biomedical GmbH, Germany). A 23Na 3D cones acquisition4,5 was performed at both sites: 4 mm isotropic, TE = 0.5 ms, TR = 100 ms, readout = 10 ms, 3 averages, 11 minutes. 1H T1 images were also acquired for registration and segmentation.
Images were reconstructed6 after applying k-space density compensation weights7, and TSC maps were generated based on calibration phantoms placed in the field of view. TSC values were obtained from white matter (WM), grey matter (GM), and CSF, based on segmentations of co-registered 1H T1 images. In addition, a manual ROI analysis was performed by two observers to assess inter-observer agreement. Coefficients of variation (CoVs) of mean values were used to assess repeatability (from scans 1 and 2 at a given site) and reproducibility (from the first scan at each site).
Example TSC maps are shown in Figure 1a, with the average WM, GM and CSF values for all subjects shown in Figure 1b. General TSC trends were consistent across sites, with mean GM concentrations slightly higher than WM; CSF values were higher and will be underestimated due to partial T1 recovery.
The repeatability and reproducibility of segmented TSC values are presented in Bland-Altman plots in Figures 2a and b, respectively. Measurements were more repeatable at site A than B, with mean repeatability coefficients of 2.4 and 4.5 mM, respectively (Figure 2a). CSF values were less reproducible than WM and GM (Figure 2b).
The repeatability and reproducibility of TSC in manually-defined ROIs are shown in Figures 3a and b, respectively, for one observer. As with the segmentation analysis, repeatability was better at site A than B, with repeatability coefficients of 2.8 and 4.7 mM, respectively (Figure 3a). The reproducibility analysis provides evidence of poorer reproducibility for higher TSC (Figure 3b). These trends were also seen in the second observers’ analysis; intraclass correlation coefficients (ICCs) calculated separately for each ROI in the first scan at each site generally indicated good inter-observer agreement (Figure 4), with ICC point estimates > 0.85 in all cases.
TSC trends in WM/GM are consistent with previous studies8,9, although GM values were slightly lower than some of the previous literature (e.g. 26.7±3.4 mM vs. 36.3±1.7 mM from Thulborn et al.10). CSF and caudate nucleus values provide evidence of poorer reproducibility of higher TSC, which may be relevant for tumors where TSC is known to be elevated11.