Magnetic resonance imaging (MRI) is a non-invasive imaging technique widely used in clinical practice. Conventional MRI is based on hydrogen nuclei because of their large gyromagnetic ratio and abundance in the human body. Notwithstanding the ubiquitous presence of ¹H-based MRI and its flexibility for multiple contrasts, other nuclei can offer insights into specific mechanisms. In recent years, the use of heteronuclear MRI, and in particular, of sodium MRI, has significantly expanded. The dynamics of sodium is crucial for several physiological processes, including neuronal excitability, synaptic transmission, and cellular energy metabolism; indeed, approximately 15% of the gray matter energy budget and 44% of the white matter ATP turnover is believed to be used only to maintain the resting potential, neglecting signaling. Overall, up to 50% of the cortical energy budget is assigned to Na⁺/K⁺ ATPase [1] and is thus needed to maintain sodium and potassium concentration gradients across the cellular membrane. For sodium, the intracellular and extracellular concentrations are approximately 10–15 mM and 140 mM, respectively [2, 3]. Dysfunctions in sodium homeostasis are associated with various neurological conditions, including multiple sclerosis (MS), Alzheimer’s disease (AD), stroke, epilepsy, and brain tumors [4]. Therefore, sodium represents a promising biomarker [5].

Sodium is the most abundant cation, providing the second strongest nuclear magnetic resonance (NMR) signal in the human body after hydrogen, but its lower gyromagnetic ratio (11.3 MHz/T), lower abundance in the human brain, and nuclear spin of 3/2 imply a sensitivity which is only 9.3% of that of the proton. These factors lead to a poor signal-to-noise ratio (SNR) [2], which is partially compensated by the use of ultra-high field strengths (7 T ad above) [6–9]. Sodium is characterized by a quadrupole magnetic moment, leading to two distinct T₂ relaxation times: a slow component of 15–40 m and a fast component of 0.5–8 ms [10–13]. The fast component accounts for approximately 60% of the total sodium NMR signal, requiring MR sequences with very short echo time (TE).

After the first pioneering sodium studies on humans and intact animals, performed in the early 1980s [14, 15], the feasibility of ²³Na-MRI in clinical settings has benefited from recent advances in MRI technology and acquisition strategies. In particular, non-Cartesian k-space sampling offers advantages in terms of spatial resolution and acquisition time [16], while ultrashort TE (UTE) or zero TE (ZTE) sequences allow the detection of the short T₂ tissue components [17]. UTE sequences begin the acquisition with a minimal delay after the excitation pulse [18], while in ZTE sequences, the sampling gradients are turned on before the radiofrequency (RF) excitation pulse. In ZTE sequences, the center of the k-space is crossed at zero echo time, and due to the switching of hardware from the transmit to receive mode, the center of the k-space is not sampled. Different strategies have been developed to overcome this problem, such as single-point acquisitions (PETRA) [19]. UTE and ZTE acquisition techniques can be customized in terms of SNR, total acquisition time, and point spread function by adapting the density of the k-space points. Examples include density-adapted radial sequence [20], 3D cones [21], twisted projection imaging (TPI) [16, 22], and FLORET [23]. Non-Cartesian sequence reconstruction methods use regridding, which involves interpolating the sampled data points in a rectilinear grid [24], and performing fast Fourier transform (FFT) or non-uniform FFT [12] on the regridded data [25].

Sodium MRI of the brain can follow different approaches according to the information of interest. The basic approach treats the brain as a single compartment and uses UTE or ZTE sequences to collect spin-density weighted (SDW) ²³Na MR data. Then, a calibration is performed in order to determine the tissue sodium concentration (TSC). However, changes in TSC can be the result of independent changes within the intracellular and extracellular compartments (IC and EC, respectively). Moreover, pathologies are often associated with changes in the volume fraction [26]. Distinguishing IC from EC sodium requires a combination of acquisition techniques and mathematical models. Most experimental approaches exploit the different relaxation times between bound sodium and free sodium [27], assuming that the intracellular component is mainly composed of bound sodium. Being generally suppression techniques, that is, techniques that selectively suppress or filter a component of the signal, they tend to suffer from low SNR and incomplete suppression.

²³Na-MRI is sensitive to physiological and pathological processes that happen during the progression of several neurological diseases [50, 51]: in fact, many brain disorders show sodium concentration anomalies associated with sodium–potassium pump dysfunction or cellular membrane impairment, allowing a sodium MRI to target specific features of the pathology. Anomalies in overall sodium are often caused by an increase in the intracellular sodium concentration or the extracellular volume fraction. Therefore, differentiating the compartments can help and improve the diagnostic information of sodium imaging.

Anomalies in sodium levels play a crucial role in the pathophysiology of multiple sclerosis [52]. The most common form of MS is characterized by a first period of relapsing/remitting (RR) symptoms that later converts into a secondary progression phase (SPMS). RRMS is primarily characterized by demyelination, resulting in impairment of action potential conduction. The demyelinated axons can experience an overexpression of sodium channels along their membranes. Albeit this process re-establishes signal transduction and contributes to clinical remission in RRMS, it is responsible for higher sodium influx, resulting in an increased energy demand to maintain resting potential and sodium homeostasis. Moreover, intra-axonal sodium accumulation contributes to axonal degeneration by reversal of the sodium/calcium exchanger that increases intra-axonal calcium (Ca), ultimately causing irreversible damage, whose accumulation leads to SPMS [53–55].

Considering the fundamental role of sodium in MS, ²³Na-MRI has been widely employed. TSC has been studied in normal-appearing white matter (NAWM), normal-appearing grey matter (NAGM), CSF, and MS lesions [55–58]. Various studies have shown an increase of TSC in lesions as well as in cortical gray and white matter across primary, secondary, and relapsing-remitting multiple sclerosis (RRMS) [55–57]. Additionally, sodium levels are linked to clinical measures of disability and impairment [57]. Studies on lesions have demonstrated that TSC is elevated in contrast-enhancing lesions, T₁ hypointense lesions, and T₁ isointense lesions, while non-enhancing lesions with a reduced apparent diffusion coefficient (ADC) exhibit TSC levels that are comparable to those in NAWM. This consistency in TSC suggests that the tissue structure may be preserved early on, prior to any breakdown of the blood–brain barrier (BBB). Thus, TSC may represent a sensitive biomarker of chronic tissue abnormalities, BBB disruption, and vasogenic edema in contrast-enhancing lesions. Normal levels of TSC in lesions with a decreased ADC may indicate that the lesions are in an early stage [10, 59, 60].

The well-demonstrated increase of TSC in MS can be driven by intracellular sodium accumulation, expansion of the extracellular space, or both [57]. Strategies based on single-quantum, inversion recovery, and triple-quantum filtering have been used to disentangle the intracellular contribution [8]. Findings suggest that intracellular sodium concentration can give complementary information about MS lesions and inflammatory processes. Petracca et al. [8] found an increase in ISC in both WM and GM of RRMS patients; a decrease of the intracellular volume fraction was observed in several clusters in WM and GM, while changes at the whole-brain scale were significant only in WM. Interestingly, lesion burden and disability were correlated to TSC and intracellular volume fraction but not to ISC, suggesting that the latter is a consequence of early neuro/axonal metabolic dysfunction, while the former reflects tissue loss [8]. Studies on lesions have demonstrated a different intracellular concentration between hyperacute, acute, and chronic ones [61, 62], suggesting the sensitivity of ISC to reveal inflammatory mechanisms during the first stage of the disease (see Table 1).

Alzheimer’s disease is characterized by extracellular deposits of β amyloid peptides (Aβ). Postmortem studies have shown a correlation between Aβ accumulation and ion imbalance [69–71]; disruption of Na⁺ homeostasis is induced by Aβ oligomers and is responsible for neuronal network destabilization in the early stage of the disease [71]. TSC was found to be high in AD patients, suggesting that sodium imbalance during the progression of the disease might have a key role in energy failure [72, 73]. However, recent ex vivo studies have shown that sodium increase in brain tissue does not correspond to an increase in the CSF, indicating that anomalies can be associated with the intracellular pool [69]. However, no direct evidence exists for this hypothesis so far.

Stroke is one of the most common causes of death and the leading cause of long-term disability. During a stroke, the blood flow in the brain is compromised, resulting in a breakdown of cellular energy production, followed by the piling of sodium and calcium in the intracellular compartment. A measure of intracellular sodium would have the benefit of being related to the direct pathological consequence of stroke. A study on non-human primates has shown that the duration of brain ischemia can be associated with a threshold on the TSC beyond which tissue reperfusion would be of little benefit [74]. Studies on humans have demonstrated how sodium can be considered a promising biomarker for tissue viability and cell integrity, and how its integration into clinical practice can have an impact on future diagnosis and treatments [75–77]. It is likely that knowledge about IC sodium would help personalized decisions on reperfusion.

Epilepsy is characterized by recurring seizures, causing a large inflow of sodium. Intracellular ²³Na-MRI is thus an obvious biomarker to quantify the physiological effects of seizures. Moreover, cell shrinkage and cerebral atrophy have been reported in epileptic areas [78, 79]. Both phenomena are associated with an increase in the extracellular volume fraction, resulting in an increase of TSC in the epileptogenic zone [80]. An increase in TSC in epileptogenic zones suggests intracellular sodium accumulation, even in interictal periods [81], and demonstrates long-term changes in sodium levels [80, 82]. A decrease in TSC was observed in a patient who suffered non-convulsive seizures during the scan, suggesting transient neuronal swelling and reduction of extracellular volume fraction [82]. Some recent studies [80, 83] investigated the ratio between T₂*_short and T₂*_long, showing an increase of the short component mainly localized to epileptogenic zones, whereas TSC was increased in all regions. These results suggest that a differential weighting of intracellular/extracellular sodium can be sufficient to highlight specific pathophysiological phenomena.

Traumatic brain injury (TBI) is the result of acute events in which an external force damages the brain. TBI causes stretch-induced damage to the cell membranes along the axons, causing mechanical damage and mechanoporation [84], triggering ionic and proteolytic cascades, and ultimately resulting in disruption of ionic homeostasis. Sodium imbalances following mild TBI are associated with patient outcomes [85]. Surprisingly, a decrease of TSC in TBI was reported [84, 85], contrary to what could be expected. A recent longitudinal study [86] has investigated TSC levels 3 months after the injury, when no significant variations of TSC were found, suggesting normalization of the sodium ionic equilibrium.

Brain tumors represent another important application for ²³Na-MRI. The most aggressive are characterized by rapid cell proliferation and angiogenesis: both of these factors are linked to a reduced Na⁺-K⁺-ATPase activity [87], resulting in an increase of the intracellular sodium concentration. Cellular proliferation can also be associated with an increase in extracellular volume fraction [88]. The level of Na⁺ in malignant tumors is higher than in healthy tissues [46, 89–95]. IC sodium is thus a potential biomarker for cancer staging. Moreover, EC volume fraction is expected to increase following treatment. Therefore, compartmentalized sodium measurements are a promising biomarker for evaluating response to treatment. Studies on compartments have demonstrated a decrease in the signal associated with intracellular sodium (see Table 1) [30, 48, 64, 66, 96]. However, Nagel et al. [6] reported an increased IR signal (associated with intracellular sodium) for tumors with high proliferation rates: a positive correlation between the MIB-1 proliferation rate and the IR signal was found, and no correlation considering TSC values was found. These results suggest the potential of compartmentalization to discriminate among different types of brain tumors [6, 67] and their progression [68].

The disentanglement of IC and EC sodium needs specific acquisition sequences and mathematical models. The scanning sequences developed for selective acquisition from a single compartment suffer from some limitations: they are based on the idea that intracellular and extracellular pools are composed of bound sodium and free sodium, respectively, and have different relaxation properties, but this hypothesis is likely an oversimplification, leading to partial or poor compartment specificity [97]. Moreover, all scanning approaches for compartmentalized sodium quantification suffer from poor SNR and a very long scanning time.

Sodium imaging is, in general, impaired by hardware and technical limitations. The challenge with RF technology for sodium MRI lies in the ability to detect small signals with very short T₂ relaxation times [98, 99], thus requiring high sensitivity and fast switching between transmit and receive modes. While most sodium coils used to be custom-built or handmade, commercial RF coils, including ²³Na/¹H birdcage coils and surface coils, are now available. Surface coils are more sensitive, but limited coverage and highly inhomogeneous B₁ field impair quantitative studies [98, 100]. In recent years, phased array sodium RF receive coils have been introduced. The efficient design of phased array receive coils and the added benefit of potential parallel imaging are attractive, but these coils necessitate correction methods for accurate sodium quantification [26, 101–103]. Acquisition strategies may also be optimized; recent studies have proposed methods to reduce scan time using postprocessing with convolutional neural networks [77] or fingerprinting to simultaneously acquire sodium density and sodium relaxation parameters and improve sodium MRI resolution [104–107].

Part of the ongoing research still relies on simulations and phantom studies. Phantoms with varying sodium and agar concentrations can simulate the relaxation behavior of compartments and aid in optimizing sequence parameters. However, further development is required regarding the quality of agar phantoms and the development of more stable alternatives to enhance optimization and sodium quantification [108].

Finally, improvements are expected from the diffusion of deep-learning-based methods for denoising or model optimization and fitting, but the relatively limited size of ²³Na datasets requires tailored solutions [109].

The first MRI studies on compartmentalization of ²³Na were presented more than 20 years ago. Because sodium compartmentalization is directly linked to cellular homeostasis, it can provide unique, relevant, and specific clinical information about tissue integrity and physiology in several serious pathologies, including cancer and neurodegenerative or neuroinflammatory diseases.

However, technical difficulties of scanning and postprocessing, as well as scanning time, currently make compartmentalized sodium imaging a tool for clinical research that is not expected to expand to clinical practice in the near future.