Brain Tumor Imaging: Diffusion, MRS & High-Field Imaging

Traditionally, brain tumor recurrence is defined as new MRI contrast enhancement apparent in follow-up imaging. This study shows that diffusion weighted MRI edge characteristics of contrast enhancing tumors show measurable differences indicative of tumor invasion prior to contrast enhancing recurrence

Diffusion kurtosis imaging evaluates the non-Gaussian diffusion pattern of water and indicates tissue structure complexity. In this diffusion study we found differences between glioblastomas and cerebral metastases, that indicate more directed diffusion in glioblastomas and a higher structural complexity in metastases.

To histopathological evaluate fiber density mapping (FDM) in glioma patients for assessment of the extent of destruction of white matter structures in the center, the transition zone and the border zone of gliomas. We correlated FDM-data and histopathological findings from 78 stereotactic biopsies of 20 glioma patients. We found a negative logarithmic correlation of fiber-density with both, % tumor infiltration and tumor cell number. For a tumor infiltration of >60% no fibers are remaining. In tumor regions with <16% tumor cells functional important fiber structures may still exists. Our histopathology-fiber-density-model may be helpful for preoperative-planning to prevent post-therapeutic neurologic deficits.

Diffusion weighted imaging (DWI) measures of apparent diffusion coefficient (ADC) is believed to reflect the level of tumor cellularity in malignant gliomas. Functional diffusion maps (fDMs) were developed to examine voxel-wise changes in ADC, then stratify voxels as either increasing ADC (indicative of necrosis or "hypocellularity"), decreasing ADC (indicative of growing tumor or "hypercellularity"), or not changing within regions of contrast-enhancement or FLAIR signal abnormality. Because the particular threshold used for voxel classification dictates the sensitivity and specificity to changes in tumor cell density, we hypothesize that a graded fDM technique that stratifies voxels into varying degrees of change, applied to the whole brain, may be useful for visualizing invading and proliferating tumor with both high sensitivity and specificity. In the current study we examine graded fDMs in 120 patients and discuss how graded fDMs can be used to detect and monitor brain tumor growth and invasion beyond the traditional malignant boundary.

Difficulties can arise in clinical practice in differentiating between primary malignant glioma and brain metastasis owing to their similar appearances on conventional MRI sequences. Although previous studies have identified differences in diffusion parameters between the tumour types, the diagnostic role of diffusion tensor imaging (DTI) has yet to be fully elucidated. We propose an application of a novel whole brain DTI segmentation algorithm in generating regions of interest (tumour and oedema) from Diffusion Colour Maps (DCMs) created using our technique. We identify differences in diffusion between tumour types, compare our method with conventional manually drawn regions of interest and propose potential clinical applications for our method.

Diffusion-weighted Imaging (DWI) is an important adjunct to standard imaging in the management of GBM patients receiving anti-angiogenic treatments. In this study, ADC values were obtained for a) areas on preprogression scans that ultimately progressed to new contrast-enhancement on progression scans (NEW_CEL), and b) new FLAIR abnormality on preprogression scans with exclusion of areas of contrast enhancement and areas that progress to new contrast-enhancement on progression scans (T2ALL_M). Results demonstrated increasing ADC values in NEW_CEL but no change in T2ALL_M in scans prior to progression. Clinical implications include interpreting new FLAIR abnormality as a consequence of anti-angiogenic treatment alone.

The purpose of this study was to use 3D ¹H MR Spectroscopic Imaging (MRSI) and diffusion tensor imaging (DTI) to develop early prognostic markers for GBM patients undergoing radiation, temozolomide and PKC inhibitor. Twenty-nine patients with newly diagnosed GBM were examined using a 3T MR scanner. Conventional anatomical imaging parameters could not distinguish between progression groups at baseline or 1 month. Parameters derived from MRSI and DTI provided information at baseline and early follow-up examinations that may be valuable in predicting the time-to-progression for patients with GBM.

Glioblastoma Multiforme (GBM) is the most common and malignant type of primary brain tumor, resulting in a median survival of approximately one year. Our study of 18 patients with GBM indicated that metabolic abnormalities more accurately reflect the underlying tumor burden. We found that the Cho to NAA index (CNI) values in the contrast-enhancing lesion (CEL) are elevated at 2 months prior to progression while having less changes in CEL volume at that time. Patients who have a CEL volume with high CNI values are more likely to progress compared with those who have with smaller CEL volume and lower CNI values. We also observed that the regions with high CNI values outside the CEL region could subsequently become enhancing.

Proton and 31P MRSI was performed on human malignant recurrent gliomas in order to provide in vivo analysis of membrane metabolism and neuronal brain damage (tNAA). Phosphorylated components in the membrane metabolism showed clear changes indicating a shift to proliferating cell fractions. While the increase in the phosphocholine/glycerophosphocholine ratio in tumor tissue did not reach significance (p=0.07) the respective ratio for the ethanolamine compound was clearly significant (p=0.02). Further, the significant increase in the inorganic-phosphate/phosphocreatine ratio hints to limited energy supply within the tumor.

The pathogenesis of tumour associated seizures (TAS), a common co-morbidity with brain tumors remains poorly understood. Glutomate has been implicated in many types of epilepsy. In a pilot study the concentration of glutamate/glutamine associated with gliomas using in vivo MRS was studied, and correlated with observed pre-operative seizures. Elevated glutamate/glutamine levels were found in the peritumoral area of tumours who experienced pre-operative seizures compared to those which did not. Due to the small sample size, we are in the process of acquiring a larger MRS and ex vivo prospective data set (N>100) to confirm these findings.

Metabolic information from 1H MRSI may aid image segmentation using DTI and so improve delineation of infiltrative brain tumours such as gliomas. NAA and fractional anisotropy (FA) are expected to decrease with tumour infiltration and loss of neuronal structure, but FA calculated from principal diffusion magnitude images is biased due to the contribution of noise. We have investigated the FA and NAA distribution in glioblastomas in comparison to simulated data that takes into account the effect of SNR on the measurement of low FA values. Our data provides evidence for diffusion anisotropy in glioblastomas in the absence of functional neurones.

MRS with Cho/Cr, mI/Cr and NAA/Cr ratios, could be more sensitive than MRI and could, in some cases, be predictive of worsening in gliomatosis follow-up. These spectroscopic changes occurred well before clinical deterioration. There is a large variability, but repetition and modelisation of spectroscopic measurements during longitudinal follow-up could allow us to diminish it and to improve gliomatosis prognostic evaluation.

Studying the relationship between MRS measures, methionine PET, segmentation and perfusion parameters could lead to better understanding of therapeutic response, especially with regard to chemotherapy and antiangiogenic molecules and in the future hypoxia modulators.
2210. Prominent Citrate Predicts Malignant Progression of Low-Grade Astrocytomas in Children

Arabhi C. Nagasunder¹, Mikhail Laskov², Albert Joseph², Ashok Panigrahy^1,3, Girish Dhall², Jonathan L. Finlay², Ignacio Gonzalez-Gomez⁴, Mark D. Krieger⁵, Marvin D. Nelson¹, Stefan Bluml<sup>1,6

1</sup>Department of Radiology, Childrens Hospital Los Angeles, Los Angeles, CA, United States; ²Childrens Center for Cancer and Blood Diseases, Childrens Hospital Los Angeles, Los Angeles, CA, United States; ³Department of Radiology, Childrens Hospital of Pittsburgh of UPMC, Pittsburgh, PA, United States; ⁴Department of Neuropathology, Childrens Hospital Los Angeles, Los Angeles, CA, United States; ⁵Department of Neurosurgery, Childrens Hospital Los Angeles, Los Angeles, CA, United States; ⁶Rudi Schulte Research Institue, Santa Barbara, CA, United States

Pediatric low-grade gliomas can either progress to a high-grade lesion or remain dormant for long periods of time. Currently, there is a need to identify markers that would allow pediatric neuro-oncologists to predict tumor progression. Our goal was to determine whether aggressive pediatric low-grade II astrocytoma have metabolic features that distinguishes them from stable grade II astrocytoma using in vivo MR Spectroscopy. We found that elevated citrate and low NAA may predict malignant progression of low-grade astrocytomas.

Sixty-three patients with EHPVO having different age groups with 47 age/sex matched controls were studied. Neuropsychological tests, MR imaging, 1H-MR Spectroscopy and blood-ammonia estimation were performed in all subjects. 40% EHPVO patients had MHE who showed significantly increased Mean diffusivity, Glx/Cr, blood-ammonia and GP T1 H in all age groups; however, mIns/Cr was significantly lower only in adults when compared to controls. Mean diffusivity positively correlated with blood-ammonia and Glx/Cr in all age groups. A significant positive correlation was observed between Glx/Cr and blood-ammonia. Increases in Mean diffusivity, Glx/Cr, blood-ammonia and GP T1 H and decrease in mIns/Cr are associated with pathogenesis of MHE in adults with EHPVO. No change of cho/Cr in EHPVO may serve as a diagnostic marker for its differentiation from cirrhosis induced MHE. A significant positive correlation among blood ammonia, Glx/Cr and mean diffusivity indicates that hyperammonia contributes to the generalized low grade cerebral edema.

We implemented lactate editing at 3T using BASING pulses and assessed its repeatability in phantoms and in human brain tumours in vivo to estimate the level of lactate and other metabolites . In phantoms, a coefficient of variation of 11% was achieved for lactate with SNR similar to in vivo. In tumours, lactate was detected, and there was a non-significant trend of lower metabolite concentrations in scan 2 than scan 1. Lactate editing may provide a useful means of simultaneously monitoring lactate, choline and lipids in vivo, all of which are of interest in tumour progression and response to treatment.

Grading of non-enhancing supratentorial gliomas and neuronal-glial tumors (NEGNGT) is a diagnostic dilemma on conventional MR imaging as 45% of these tumors could be malignant. We retrospectively compared diffusion tensor imaging (DTI), MR perfusion weighted imaging (PWI) and MR spectroscopic imaging in preoperative grading of 50 patients with histology confirmed non-enhancing supratentorial gliomas and neuronal-glial tumors. The imaging parameters, included mean FA, mean FA ratio, maximal FA, minimal ADC, maximal rCBV, Cho/Cr and Cho/NAA which all were evaluated for both tumor groups. There were significant differences of mean FA, mean FA ratio and maximal FA between low and high grade NEGNGT(p<0.05), but no significance was found for the other parameters. ROC analysis showed that the maximal FA value had a higher sensitivity and specificity than the other parameters to differentiate between low and high grade NEGNGT. This result indicates that maximal FA may be the best adjuvant tool to help differentiating between low and high grade in non-enhancing supratentorial gliomas and neuronal-glial tumors.

This study addresses the inability of conventional imaging to delineate infiltrative tumor regions, which causes the tumor to recur within the treatment volume. These regions have normal cellularity due to diffusely infiltrating cells and cannot be located based on increased cellularity with choline and ADC. To define an MR profile of an infiltrative tumor region HR-MAS spectroscopy of image-guided biopsies in newly diagnosed and recurrent GBM were analyzed and evaluated relative their spatial location within the tumor, pathological measures of its paired sample and diffusion measures derived from the biopsy location. These results will be presented and discussed.

Malignant gliomas are highly vascularized, which makes them a target for new anti-angiogenic agents. MRI therapy monitoring will require more sophisticated methods than measuring the extent of contrast enhancement, because these agents change the blood-brain-barrier. Using the advantage of 7 T, direct imaging of the tumor vasculature becomes feasible. In our newly developed clinical imaging protocol, high-resolution T2w and T1w images show the internal morphology of the lesion, whereas TOF-MRA and SWI visualize the arterial and venous intratumoral vasculature. Automatic co-registration of MRA and morphology proved to be a simple, fast and reliable method to evaluate tumor vascularization.

MRI DSC techniques offer an efficient way to characterize brain perfusion properties. However, the use of low-molecular weight gadolinium based contrast agents (Gd) can introduce a significant confound into standard DSC analysis if extravasation is extensive; as often is the case for brain tumors. The purpose of this study was to investigate the use of an ultra-small paramagnetic iron oxide (USPIO) compound to estimate brain tumor blood volume in human subjects. To accomplish this goal, DSC acquisitions were performed using both Gd and USPIO based contrast agents in eleven human subjects at 3T and 7T.