Neuroimaging and Cartography: Mapping Brain Tumors

Cartography: The art or technique of making maps.Map: A representation, usually in a plane surface, of a region. To plan or delineate in detail, to explore or make a survey.

Although MR images are currently of exquisite resolution, as neuroradiologists we all acknowledge that confidently mapping the margins of primary brain tumors is very difficult. Malignant gliomas tend to extend beyond their enhancement into areas of high T2 signal that most of us initially assume to represent edema. This behavior is typical of primary gliomas, and at our institution, if proton MR spectroscopy (MRS) shows elevated choline (Cho) outside of the area of enhancement, we generally nearly always exclude the diagnosis of a solitary metastasis (these are well-marginated lesions without neoplastic infiltration of edema). As neuroradiologists, our primary role is that of mapping the extent of a tumor and then predicting the histologic findings. The former is important because the success of many current and future therapies will depend on it. Currently, the extent of the initial research based on information obtained from imaging studies is still the factor that contributes the most to the patient's prognosis (1). Apart from MR imaging, what can we do to map tumors?

Magnetization transfer rates (MTR) show higher values in higher-grade gliomas than in edema. Radiation necrosis shows low MTR but in a similar range as tumors (2). Thus, MTR cannot be used to map treated tumors. We attempted to use apparent diffusion coefficients (ADC) to separate tumors from edema and normal tissue (3). In our experience, ADC could not reliably distinguish between tumor and edema. It is possible that diffusion tensor imaging will show altered anisotropy in areas of tumor infiltration, but preservation of anisotropy in regions of edema where the white matter tracts are not destroyed. MRS is yet another technique that may be used to map tumor margins. Up until now, the major limitations of MRS are that 1D and 2D techniques suffer from partial averaging effects and limited spatial resolution.

In this issue of the AJNR, Dowling et al (page 604) take an important step in establishing MRS as a promising technique to map tumor extension. They obtained 3D MR spectra in 28 brain tumors and correlated the histologic findings from specific sites to the corresponding voxels. From their results we learn the following:

1. If N-acetyl aspartate (NAA) is normal, the tissue is normal. Conversely, abnormal NAA correlates with abnormal tissue regardless of the underlying histologic findings.

2. When a lesion contained a Cho level that was larger than NAA and larger than normal Cho, tumor was always present.

3. The level of Cho correlated with the percentage of tumor in the specimen.

4. Non-detectable to near-normal levels of NAA and Cho were seen in areas of gliosis/necrosis.

5. Areas of similar appearance and enhancement may show different spectra and different histologic findings.

On the basis of these data, it seems that MRS is a worthy method that may be used to map tumor margins, identify areas with the highest malignancy, and serve to guide biopsy, identify residual tumor after surgery, and identify recurrent tumor after treatment.

Despite these exciting possibilities, the study by Dowling et al also has some limitations. Their data were correlated retrospectively, and it is not clear if in an a priori analysis of MRS will perform similarly (although I have the impression that it will). The studies loaded into the imaging-guided surgery device were not the same as those used to guide the placement of the MRS grid. Indeed, in some patients, considerable time elapsed between biopsy and MRS; thus, the correlation between these two may not be entirely accurate. Some authors believe that neuronavigation systems do not identify deep tumor reliably (4). This is attributable to deformation of the brain after craniotomy and shifting of the structures. It is, thus, possible that brain shift may also have introduced some degree of sampling error in Dowling's patients. Not all biopsies were obtained from the most abnormal-appearing spectra; therefore, the histologic grade of some lesions may have been underestimated. The authors clearly point out that volume of tissue obtained comprises only a small percentage of the size of the MRS voxels. At present, the lack of commercially available 3D MRS packages also limits the widespread use of the technique. The authors correctly point out that MRS and MR imaging may be performed in one setting. It is unclear if the MRS studies were guided by and obtained after contrast administration. They observed that Cho levels tended to be higher in tumors grade II and III but slightly lower in glioblastoma multiforme. This is counterintuitive and is probably related to partial volume effects.

In very few years, MRS has matured and become clinically feasible in most MR units. Because many other techniques, such as MTR and ADC maps, are limited in the mapping of brain tumors, careful prospective studies using MRS are needed. The study of Dowling et al supports my impression that MRS provides good tissue characterization and is very helpful in the mapping of brain tumor margins. We need to incorporate the information obtained from MRS into imaging-guided surgery devices to obtain full benefit of these exciting techniques. Detailed mapping of tumor margins are needed before robotic brain surgery becomes clinically feasible. Mapping of tumors will also he critical for the placement of intralesional chemotherapy, delivery of viral vectors or stem cells, and guiding of stereotatic radiosurgery. Because the brain cannot adequately remove necrotic detritus, mapping of necrosis is also important. Identification of exact locations of brain damaged by tumor may lead to implantation of neuronal and glial precursor cells, which may minimize deficits. It seems that, as neuroradiologists and brain cartographers, we are moving in the right direction.