Considering the lower gyromagnetic ratio and the lower number of nuclei in a distinct chemical bonding for metabolites (see above), the inherent sensitivity for a compound like PCr, which is visible in ¹H and ³¹P MRS, should be about 20 % of its ¹H sensitivity. This simply requires a ³¹P voxel volume of five times the ¹H-voxel volume or, assuming isotropic resolution, an increase in matrix grid size by almost 75 %. When ¹H MRSI and ³¹P MRSI are obtained with the intention to combine data, a reasonable geometry would match the slice thickness and fit four ¹H voxels into one ³¹P voxel. On top of the larger voxel size, an increased spreading of signal into adjacent voxels caused by the point spread function, which describes the blurring of a point due to coarse k-space sampling, has also been considered. Figure 3 demonstrates the effect: It shows a T2-weighted anatomical slice with a tumor in panel a. Panel b depicts the same slice; however, by digital filtering a blurring was implied which mimics a resolution typical to a ¹H MRSI data set obtained with a 16 × 16 matrix at 240 mm² FOV. Panel c shows this slice at a resolution corresponding to an 8 × 8 × 8 matrix for data acquisition followed by digital resolution enhancement. Obviously, the T2 enhancement marking the tumor is still clearly marked at the ¹H MRSI resolution but tends to be blurred across the entire brain at the rather poor k-space sampling employed for ³¹P MRSI.

The unique feature of ³¹P MRS is the ability of the method to detect the pH values in tissue. The effect is based on the pH-dependent equilibrium between HPO₄ ²⁻ and H₂PO₄ ⁻ with the signal position of inorganic phosphate providing a weighted average of both ions (Prichard et al. 1983). Many studies observed a shift of the Pi signal toward a more alkaline environment (Negendank 1992). With the assumption that this signal originates mainly from intracellular phosphate (Stubbs et al. 1992), the ³¹P MRS data indicate a reversed pH gradient in tumor tissue (more acidic extracellular) with respective consequences for tumor metabolism and drug efficiency (Stubbs et al. 1994). Figure 4 shows typical ³¹P MRS spectra from normal-appearing tissue (black) and the tumor (red). The insert at the upper right corner represents the extended Pi region. For the tumor spectra, the Pi signal exhibits a shoulder to the left (lower frequencies or higher pH) and a shift in this direction.

We want to stress here that partial volume effects prevent from a thorough analysis of tumor pH, leaving the open question whether the lower pH fractions are due to healthy tissue adjacent to the tumor or result from extracellular Pi. A recent publication at least identifies the averaged Pi position as a sensitive marker for treatment (Hattingen et al. 2011).

While the tCho signal in ¹H MRS just reflects an overall concentration of choline compounds, its main components GPC and PCho can be discriminated with ³¹P MRS. However, for MRSI data the quantitative analysis in terms of tissue concentrations is severely hampered by partial volume effects as described above. The poor point spread function related to the standard 8 × 8 matrix for data acquisition is causing a sensitivity profile with enhanced signal intensity in the center of the brain. Correction by dividing by tissue content obtained from anatomical MRI after appropriate filtering (see Fig. 3) may account for this, but experimental data on recurrent brain tumors show that the ratio of PCho/GPC, which provides an inherent correction for this, gives better results for comparing tumor to healthy tissue and monitoring treatment (Hattingen et al. 2011; Hattingen et al. 2013). A further advantage of ³¹P MRS is the ability to observe the ethanolamine analogues to the choline membrane metabolism adding more variables for determining tumor-specific biochemistry.

Metabolite signals from Pi, ATP, and PCr can easily be detected in the ³¹P MRSI data allowing an estimation of changes in energy metabolism by comparing the ratio of the high-energy phosphates (ATP, PCr) to Pi. Most publications report a consistent drop in this ratio (Negendank 1992; Hattingen et al. 2011; Hubesch et al. 1990; Rutter et al. 1995; Maintz et al. 2002) indicating limited energy supply in tumor tissue.

De Graaf et al. reviewed the unique role of ¹³C MRS in measuring brain metabolism in vivo (de Graaf et al. 2003) while Ross et al. described preliminary clinical experience in the same journal (Ross et al. 2003). However, due to the poor sensitivity, the in vivo application of ¹³C MRS to focal lesions in the human brain is even more hampered by limited spatial resolution and partial volume effect than ³¹P MRS. At least there is one case study (Wijnen et al. 2010) reporting on a malignant glioma. In this study, ¹³C labeling of lactate and glutamate was observed during an intravenous infusion of 100 % enriched [1-¹³C]-glucose solution (20 %, w/w). The authors compared 54 × 32 × 29 mm³ voxels, one placed onto the tumor and the other one in the contralateral hemisphere, monitoring the increase in labeled lactate and glutamate. As expected, only glutamate labeling was observed in the normal-appearing tissue while the tumor voxel showed increased lactate labeling and diminished labeling of glutamate. Based on the glutamate labeling in the reference voxel and correcting for the partial volume effect in the tumor voxel, the authors concluded that no glutamate labeling occurs in tumor tissue, indicating no oxidative phosphorylation in tumor cells in accordance with the Warburg effect (Warburg 1956). A more recent study performed in situ ¹³C labeling with uniformly labeled ¹³C glucose during surgery (Maher et al. 2012). Perchloric acid extracts from excised tumor specimens were then analyzed by high-resolution NMR spectroscopy. This study enrolled 11 patients with different types of tumors (including metastasis). Despite the variability in tumor type, the authors observed some common features in carbohydrate metabolism: (1) glucose is entering the TCA cycle indicating mitochondrial oxidation of glucose; (2) formation of glycine, glutamate, and glutamine from glucose; (3) in tumor tissue, the TCA cycle oxidizes alternative carbon sources indicating anaplerosis probably due to a high demand for other cell metabolites. In summary, these data highlight the potential of ¹³C MRS to provide tumor-specific biochemical information, but a noninvasive in vivo protocol for monitoring focal lesions is not available yet.