The potential of
23Na-MRI for brain imaging in oncological and non-oncological applications remains to be ultimately determined [
29]. To further implement
23Na-MRI in clinical neuroimaging, particularly in the setting of oncological follow-up, certain fundamental requirements are to be fulfilled. One of these is the intra-individual reliability and reproducibility of brain
23Na-concentrations on consecutive MR examinations. Little published data are available regarding the
23Na-concentrations of healthy gray and white matter, but the values of the current study are in the range described in the literature [
20,
21,
30]. This study showed no significant asymmetries in the
23Na-concentrations of several predefined areas of the human brain, other than between the left and right caudate head. These differences could be attributed to partial volume effects of the immediately adjacent cerebrospinal fluid (CSF). The relatively narrow head of the caudate nucleus impeded the manual ROI placement, which could explain these differences. One way to potentially solve this problem is by normalization of the data to a reference brain data set, as done in psychiatry studies for hippocampus volume measurements and the use of semi-automatic segmentation tools. Statistically significant intra-individual differences were neither seen between two consecutive MR exams nor between MR exams separated temporally by 8 days. This supports the repeatability and reproducibility of cerebral
23Na-MRI. However, per-patient differences in sodium concentration were as high as 10%, depending on the anatomic region. This potential bias should be considered in any further longitudinal or follow up study. Nevertheless, future studies may benefit from an intra-individual longitudinal approach, since inter-individual differences at a single time point can reach over 30%. Another important consideration for the clinical implementation of brain
23Na-MRI would be the development of a simple evaluation method for
23Na-MRI imaging with sufficient inter-reader reliability. The inter-reader reliability in this study was moderate but not perfect; however, this may be explained in part by the differences in clinical
23Na-MRI experience between the readers (6 vs. 2 years). As previously mentioned, a semi-automatic segmentation tool could facilitate this assessment and decrease inter-reader variability. In general, several future avenues may benefit from a broader clinical implementation of
23Na-MRI. Accuracy and quantification of
23Na concentrations between different research facilities, hospitals, and techniques (e.g. UTE, radial or cones scheme [
31]) should be confirmed. Double-tuned coils seem to be the ideal instrument for
23Na-imaging enabling the appropriate fusion of
1H- and
23Na-images. Although the reliability of the current technique appears to be sufficient, this potential (small) for bias should be taken into account within further studies. While intra-individual comparison seems acceptable for assessment of
23Na concentration, biases may arise due to natural inter-individual differences, and further follow-up studies are necessary. Finally, future semi-automatic segmentation tools as proposed could help to partially overcome biases due to inter-reader interpretative differences.