Tuning in on Tumor Activity with Proton MR Spectroscopy

Proton MR spectroscopy's role in the clinical evaluation of human brain tumors has been receiving increased attention from neuroradiologists, neurosurgeons, and radiation and medical oncologists. Good-quality proton MR spectra can be obtained on most clinical 1.5-T MR imaging systems fitted with commercially available automated software that allows acquisition of single-volume MR spectra. More recently, multi-volume proton MR spectroscopic techniques have become available that allow exploration of multiple-volume elements as small as 0.5 cc within a slice as quickly as 10 minutes. Additionally, in multi-volume MR spectroscopic studies, processing software programs are available that can display the relative level of the proton metabolites within the spectroscopic voxels. Zones of colors of varying intensity or shades of gray can be overlaid onto an MR image to show the distribution and level of the metabolic group within the anatomic image slice, known as either spectroscopic or chemical shift imaging. Because of these advances, we and others have integrated the proton spectroscopic technique into the routine clinical evaluation of brain tumors, because it provides greater information concerning tumor activity and characterization of the tumor tissue than what is possible with contrast-enhanced MR imaging techniques alone.

Clinical studies have shown the value of proton MR spectroscopy for the differentiation of recurrent or residual brain tumor activity from necrotic and cystic tumor processes (1–3). The differentiation of these processes is based on comparison of the proton spectral peak patterns and peak area determinations from normalized spectra for the intracellular brain metabolites choline (3.2 ppm), creatine (3.0 ppm), N-acetyl aspartate (2.0 ppm), lactate (1.3 ppm), and lipids (0.8–1.5 ppm). Increases in choline levels relative to creatine and N-acetyl aspartate measured on pretreatment MR spectra have been shown to correlate with the proliferative capacity of gliomas (4). Medium-to-high choline levels, relative to creatine, have been used as a marker for the presence of actively proliferating tumor cells, whereas decreases in the overall levels of choline, creatine, N-acetyl aspartate, and increases in lipid/lactate proton resonances between 0.8–1.5 ppm indicate necrotic processes. These changes in brain metabolite levels/patterns have been used to monitor and assess the effects of therapy on brain lesions (5).

In this issue of the AJNR, Martin et al (page 959) have extended the use of proton MR spectroscopy in assessing brain lesions. In this article, the authors have used proton MR spectroscopy in conjunction with MR imaging to intraoperatively guide them in selecting areas for biopsy within the tumor that have the greatest activity as judged by the level of choline relative to the disease-free area of the brain. The rationale for the use of proton MR spectroscopy to target the region for biopsy is that the selection of the biopsy target normally is based on the tumor's anatomic appearance and its enhancement properties. However, in patients who have heterogeneous lesions or who have been treated with radiation, the CT or MR imaging findings “may be insufficient for defining an optimal target for pathologic assessment”. Addition of a technique that will give information about tumor metabolic activity at the intended biopsy site clearly is desirable to help define the optimal site from which to obtain a needle biopsy specimen.

A proton turbo spectroscopic imaging technique was used in conjunction with non-contrast MR imaging to select and position the needle sites for biopsy in 26 patients. Of these patients, 16 had received prior radiation treatment for their tumor, nine were being assessed for the first time, and one had undergone a hemispherectomy 11 years prior to the present study. Both the spectroscopic imaging data sets were used to determine an appropriate target site. Metabolic images were created that could be overlaid onto the anatomic images. The choline spectroscopic imaging maps were used extensively to locate focal regions of relatively high levels of this metabolite. After selecting the target from the combined spectroscopic and anatomic image and creation of a burrhole, the patient was repositioned in the MR unit and the introduction of the biopsy needle into the brain was monitored using real-time snapshot MR images to position the needle precisely at the desired target. The patient was then removed from the magnet and the tissue was harvested by means of the needle.

Of the 26 patients, only 17 manifested focal MR spectra with regions of increased choline and had histologically confirmed tumor. In the remaining nine patients, proton MR spectroscopy did not show any regions that had major increases in choline levels and had spectra suggestive of necrosis. Of these nine patients, five had histologically proven necrosis, whereas in the remaining four patients (two glioblastoma multiforme, one lymphoma, and one germinoma), the presence of tumor along with necrosis was histologically confirmed. In one histologically confirmed glioblastoma multiforme with a choline level less than or isointense to the reference volume, the observed spectral pattern was suspicious for the presence of tumor (see Figure 7 [page 967]). However, even though the investigators felt that this area was suspicious, they felt that they could not include this in their MR spectroscopic/histologic positively correlated patient data set. We consistently find similar MR spectral patterns among patients who have glioblastoma multiforme in which volume contains a large level of necrosis (both treated and untreated) and classify volume containing such spectral patterns as positive for active tumor. If this patient is included, in which the proton MR spectroscopic findings correlated with the histologic findings, then only three of 26 or approximately 10% of the areas that underwent biopsy gave false-negative results.

In this preliminary study, it appears that proton MR spectroscopy in conjunction with MR imaging will be highly useful in defining areas for the stereotactic biopsy of brain lesions. Nonetheless, despite the encouraging results from this study, there are several limitations. The major one is that most institutions do not have intraoperative MR suites in which these studies can be performed. Another limitation is that the spectroscopic method used in this study was a 2D and not a 3D MR spectroscopic technique (3). Thus, not all of the tumor could be examined. A 3D MR spectroscopic technique would have allowed visualization of metabolic activity throughout the entire tumor and surrounding regions. Additionally, if 3D metabolic spectra can be obtained, the possibility exists for the incorporation of this data set in 3D MR/CT imaging-guided surgery devices now used at many institutions to obtain fused anatomic and spectroscopic images that can be used for both biopsy and surgical planning. This may obviate the need for these studies to be performed in a dedicated MR intraoperative suite.