While the morphologic and functional properties of normal or neoplastic tissue in the prostate gland can be delineated by MR imaging methods, in vivo proton MR spectroscopy (¹H-MRS) and spectroscopic imaging (¹H-MRSI) investigate its chemical composition noninvasively and thus yield further insight into prostate metabolism and the metabolic alterations caused by cancerous infiltration. Also, metabolic changes in the course of radiotherapy can be monitored, and the response of PCa to treatment may be assessed by MRS and MRSI. Of the metabolites present in the prostate gland, citrate (Cit), creatine/phosphocreatine (Cr), choline-containing compounds (Cho), and polyamines (PA) have sufficiently high tissue concentrations (above 1 mmol/kg) to be detected by MRS at the magnetic field strengths used for in vivo examinations. Cit is produced by oxidative phosphorylation within the citrate cycle and is extensively stored in healthy prostate tissue mainly in bound form as zinc citrate (Costello et al. 2005). It has a ¹H-MRS resonance at 2.65 ppm (chemical shift relative to tetramethylsilane as reference) arising from the non-equivalent methylene protons (CH₂) which are strongly coupled to form two almost overlapping spin doublets. This leads to a characteristic 4-peak spectral pattern at 3 T with two equally high central peaks spaced by 8 Hz and two smaller “outer” satellites each at 16 Hz distance (corresponding to the J coupling constant of Cit) from the central peaks. However, full resolution of all 4 peaks is achieved in vivo only with excellent magnetic field homogeneity, while in many prostate MRS acquisitions at 3 T, the two central constituents appear as a single broadened peak, which is a general issue at the lower field of 1.5 T. Also, the three methyl resonances of Cho, PA, and Cr covering the spectral range of 3.21–3.03 ppm are often not completely resolved, especially at 1.5 T, and therefore, usually the summed intensity (tChoCr) of all peaks in this frequency range is compared to the total area under the citrate components, yielding a metabolite ratio, e.g. called tChoCr/Cit.

Although being itself a complex overlay of several constituents such as glycerol phosphorylcholine (GPC), phosphorylcholine (PC), acetylcholine (ACho), and free choline, the intensity of the Cho peak centered at 3.21 ppm in the proton spectrum of prostate tissue may serve as a biomarker for the detection of malignant disease, just in analogy to the PSA value, with strongly increased Cho level or ratio Cho/Cit being suspicious for PCa (Cornel et al. 1993). As especially the phospholipids GPC and PC are key components released in cell membrane turnover, extensive cell proliferation as it is found in malignant tumors is often accompanied by a characteristic elevation of the choline peak in the ¹H-MR spectrum. Simultaneously, the accumulation of citrate is inhibited in cancerous prostate tissue, and the MRS intensity of the citrate peaks decreases (Costello and Franklin 1997). Therefore, both effects add up to increase the metabolite ratio Cho/Cit. In contrast, the summed area tChoCr under the Cho, PA, and Cr spectral peaks and its ratio to Cit are less sensitive to indicate PCa, because an elevation in choline levels is at least partly counterbalanced by a reduction of Cr and PA in affected tissue. Nevertheless, in cases or at field strengths with insufficient spectral separation of Cho from the adjacent PA and Cr peaks, the tChoCr/Cit ratio may still serve as a suitable marker to discriminate between benign tissue and PCa, with values <0.8 being considered as normal and tChoCr/Cit >1 as highly suspicious for malignant disease. Correspondingly, a Cho/Cit ratio below 0.5 may indicate benign hyperplasia, while Cho/Cit ratio >0.6 suspects PCa, with borderline assignment for values in between (Crehange et al. 2011). However, such thresholds have to be regarded with care as they may vary depending on the used field strength, the scan parameters (mainly TR and TE) of the MRS acquisition sequence, and the achieved spectral peak resolution. Moreover, the metabolite ratios differ between the prostate zones (with normal Cho/Cit being lower in the peripheral zone), and the cutoff values for discrimination between benign tissue and tumor have to be adjusted to the respective location within the gland.

While ¹H-MRS has been shown to be rather specific in the detection of PCa (89–91 % specificity), its sensitivity (75–77 %) is still inferior to other modalities (Manenti et al. 2006). One reason for missing a PCa lesion might be a too small or lacking increase of choline in tumors with only moderate cell proliferation, which is possibly associated with a lower Gleason score. Such a correlation between the Gleason score and the Cho levels in MRS has been found in some, but not in all studies (Zakian et al. 2005; Scheenen et al. 2007; Kobus et al. 2011). Also, PCa with focal size less than 1 cm is prone to be missed by MRS due to its limited spatial resolution, resulting in partial volume averaging with healthy tissue and thus yielding metabolite ratios below the chosen malignancy threshold. On the other hand, false-positive findings may be derived from high Cho signal also occurring in prostatitis, or if an intense narrow spectral peak is observed in MR spectra from the posterior parts in the basal region of the prostate, at the choline frequency of 3.2 ppm. This signal can be assigned to GPC contained in the seminal vesicles, and attention has to be paid not to mistake this “normal” GPC peak from seminal fluid with a pathological elevation of choline levels suspecting prostate cancer in that region.

The achievable spatial resolution for in vivo prostate MRS is one of the major drawbacks of this technique as the size of a tissue volume from which a proton MR spectrum is obtained by far exceeds the spatial dimensions of all other MR imaging methods and of most other imaging modalities. This is a consequence of the more than 10,000-fold lower tissue concentrations of the ¹H metabolites of interest (Cho, Cr, Cit) compared to the water protons used in MR imaging. Moreover, the intense spectral peak of water at 4.7 ppm has to be suppressed by suitable prepulses during MRS acquisition (and additionally by filtering algorithms in MRS postprocessing) to allow reliable quantification of the very small metabolite peaks, even when their chemical shift relative to water is quite large. Also, MRS signals from different molecular groups in lipids may overlay and distort the metabolite peaks, particularly of citrate, if the selected MRS volume partially includes fatty tissue. In localized single-volume (SV) ¹H-MRS, the metabolite signals are collected from within a brick-shaped tissue volume interactively placed on localizer MR images, which is selected by combining the excitation and refocusing RF pulses with magnetic field gradients for spatial encoding. This volume selection can be performed either with the point-resolved spectroscopy (PRESS) (Bottomley 1987) or the stimulated echo acquisition mode (STEAM) (Frahm et al. 1989) technique. Although the SNR of the metabolite peaks in the acquired MR spectra can strongly be improved by repeating the excitation of the volume of interest (VOI) and accumulating the MRS data (“signal averaging”), this may lead to unacceptably long measurement times if VOI sizes smaller than 2–3 cm³ are desired, even at higher magnetic fields with their inherently better SNR. Therefore, in MRS of the prostate, SV techniques are not suited for accurate localization of a tumor within the tissue as the required VOI size would extend over a much too large part of the prostate gland. SV-MRS of prostate cancer might thus only be of interest if the tumor site is already known and the time course of progression or response to therapy is to be investigated in consecutive examinations.

By application of the MRSI technique (often also called chemical shift imaging (CSI)), a 2D or 3D grid consisting of a multitude of smaller voxels can be used to collect MR spectra from each of these voxels. To achieve this, similar to the principles of MR imaging, slice selection is performed by gradient switching during RF excitation, and additional phase encoding is applied for in-plane localization. In contrast to the MR imaging of water protons, however, it is not possible to use frequency encoding to acquire a complete row of the image following a single excitation, because the different frequency components of the detected MRS signal are already linked to the spectral information on the metabolites of interest. Therefore, phase encoding for two spatial directions (or even three in 3D-MRSI) has to be used, and as a consequence, m × n spin excitations spaced by the repetition time TR have to be performed to acquire the desired matrix of m × n MR spectra from all 2D grid voxels. Fortunately, the SNR in these spectra is not determined by the spin signal only from the corresponding small voxel, but from the total MRSI grid which above all is sampled m × n times and accumulated. In this way, in-plane voxel sizes around 1 cm² or below can be utilized with sufficient metabolite SNR in 2D-MRSI of the prostate, with typically 10 × 10–16 × 16 voxels over a grid extension of 8–12 cm. Because a field of view (FOV) of this size is completely surrounded by body tissue, back-folding of spin signal from outer structures would happen, like in MR imaging with such a small FOV, if only the 2D phase encoding was used for in-plane volume selection. Therefore, additional PRESS or STEAM localization of a VOI smaller than the margins of the phase-encoded FOV but completely covering the prostate gland has to be applied to avoid artificial signal contribution in the spectra of the MRSI voxels, especially from lipids. Suppression of such “outer volume” signals can also be achieved by multiple regional presaturation bars closely adapted to the individual prostate shape. In most cases, the cranio-caudal size of the prostate is too large to be entirely included in a 2D-MRSI acquisition with a slice thickness of 1–2 cm, and a 3D phase encoding scheme with at least three axial slices has to be used then. Fig. 4 shows an example for the image-guided planning of such a 3D-MRSI acquisition (TR/TE 1200/135 ms) with display of the 10 × 10 × 3 voxel grid (red), the PRESS-localized VOI (green frame), and a selection of spectra (yellow-framed voxels) from the peripheral and from the central zone of the prostate within the VOI. In this examination performed before radiotherapy, a high Cho peak and low Cit level indicating extensive tumorous infiltration are present in the spectra from the posterior region of the prostate, with accentuation in the right glandular lobe and apical. Just as the other techniques of multiparametric MR imaging, also MRS and MRSI of the prostate profit from the signal gain achievable by the use of an endorectal RF coil for signal detection, and its application allows to further decrease the minimum voxel size. While at magnetic fields of 2 T or less, the application of ERCs for prostate MRSI is mandatory, at 3 T the inherently higher MR signal and the stronger sensitivity to susceptibility artifacts caused by the endorectal placement might balance out the advantages of such coils. Therefore, also considering the signal increase gained by recent progress in MR detection sensitivity by digital RF chains and coil development, the sole use of external surface coils will at 3 T supply sufficient SNR and spatial resolution for MRSI, combined with more patient comfort.