Correlation ofMRI-Derived Apparent Diffusion Coefficients in Newly Diagnosed Gliomas with [F]-Fluoro-L-Dopa PET: What AreWe Really Measuring withMinimumADC?

BACKGROUND AND PURPOSE: There is significant interest in whether diffusion-weighted MR imaging indices, such as the minimum apparent diffusion coefficient, may be useful clinically for preoperative tumor grading and treatment planning. To help establish the pathologic correlate ofminimumADC, we undertook a study investigating the relationship betweenminimumADC andmaximum FDOPA PET uptake in patients with newly diagnosed glioblastoma multiforme. MATERIALS AND METHODS: MR imaging and FDOPA PET data were acquired preoperatively from 15 patients who were subsequently diagnosedwith high-grade brain tumor (WHOgrade III or IV) by histopathologic analysis. ADC and SUVR normalized FDOPAPETmapswere registered to the corresponding CE MR imaging. Regions of minimum ADC within the FDOPA-defined tumor volume were anatomically correlated with areas of maximum FDOPA SUVR uptake. RESULTS: Minimal anatomic overlap was found between regions exhibiting minimum ADC (a putative marker of tumor cellularity) and maximum FDOPA SUVR uptake (a marker of tumor infiltration and proliferation). FDOPA SUVR measures for tumoral regions exhibiting minimum ADC (1.36 0.22) were significantly reduced compared with those with maximum FDOPA uptake (2.45 0.88, P .0001). CONCLUSIONS: Therewas a poor correlation betweenminimumADC and themost viable/aggressive component of high-grade gliomas. This study suggests that other factors, such as tissue compression and ischemia, may be contributing to restricted diffusion in GBM. Caution should be exercised in the clinical use of minimum ADC as a marker of tumor grade and the use of this index for guiding tumor biopsies preoperatively. ABBREVIATIONS: CE contrast-enhanced; FDOPA 3,4-dihydroxy-6-[F]-fluoro-L-phenylalanine; GBM glioblastomamultiforme; SUVR standardized uptake value ratio; WHO World Health Organization Despite the recent advances in neurosurgical resection techniques and adjuvant therapies, the median survival for patients diagnosed with glioblastoma multiforme remains poor at 15 months. This devastating outcome can be attributed to 2 major factors: 1) the limitation of currently used diagnostic imaging, in particular MR imaging technology, to provide clinically relevant information about tumor proliferation and physiology; and 2) the failure of current therapies for targeting extremely invasive proliferating tumor cells, in many cases anatomically isolated from the main tumor mass. Routine contrast-enhanced MR imaging plays a pivotal role in the planning of treatment strategies. However, CE MR imaging only detects dysfunction of the blood-brain barrier, which, in many cases, may not correspond with tumor proliferation or other molecular events. Many of these limitations have been overcome by using PET molecular imaging technology with amino acid– based tumor cell tracers, such as methyl-C-L-methionine and 3,4-dihydroxy-6-[F]fluoro-L-phenylalanine. Such tracers have been shown to be superior in the diagnostic assessment of patients with brain tumor compared with CE MR imaging and 2-deoxy-2-[F]fluorodeoxyglucose PET. Due to the invasiveness and logistic constraints associated Received February 21, 20102; accepted after revision July 6. From the Centre for Clinical Research (S.R.), Centre for Medical Diagnostic Technologies in Queensland (Y.G., S.C.), and Discipline of Medical Imaging (A.C.), University of Queensland, St Lucia, Brisbane, Australia; and Department of Radiation Oncology (M.F.), Queensland PET Service (P.T.), Australian e-Health Research Centre (P.B., O.S., S.R.), Commonwealth Scientific and Industrial Research Organization; and Department of Medical Imaging (A.C.), Royal Brisbane and Women’s Hospital, Herston, Brisbane, Australia. This work was funded by the National Health and Medical Research Council (grant 631567) and a Queensland Government National and International Alliances Research Program grant. Please address correspondence to Stephen Rose, PhD, Centre for Clinical Research, Royal Brisbane and Women’s Hospital, Brisbane 4029, Australia; e-mail: [email protected] Indicates open access to non-subscribers at www.ajnr.org http://dx.doi.org/10.3174/ajnr.A3315 AJNR Am J Neuroradiol ●:● ● 2013 www.ajnr.org 1 Published October 18, 2012 as 10.3174/ajnr.A3315 Copyright 2012 by American Society of Neuroradiology. with routine clinical PET imaging, significant interest has been directed toward the clinical development of diffusion-weighted MR imaging indices, such as the apparent diffusion coefficient, for preoperative tumor grading and treatment planning. Because ADC provides quantitative information about tumor physiology on a macroscopic scale, the technique shows promise for aiding image-guided therapy, especially within biologically heterogeneous tumors such as GBM. However, the clinical utility of the technique for assessing tumor grade has yet to be established. Numerous studies have reported mixed findings regarding the use of ADC to measure tumor grade. The rationale behind the use of ADC indices is based on the premise that tumor cellularity is inversely related to the ADC (ie, tumoral regions with low, sometimes termed “minimum,” ADC correspond to areas of high cellularity). One study has shown a significant correlation between minimum ADC and a measure of proliferation index (Ki– 67) for astrocytic tumors in general; however, this correlation was not significant for GBM alone. Because no study has yet performed minimum ADC–image-guided biopsy for histologic confirmation, this assumption remains highly speculative. Despite this constraint, a significant correlation has been established between the minimum ADC and reduced patient survival. A recent study aimed at investigating the relationship between minimum ADC and FDG-PET reported correlations between regions of low ADC and enhanced FDG uptake in patients with GBM. However, the clinical interpretation of FDG uptake in primary brain tumors is complex due to high background glucose metabolism, especially within the cortex, and false-positive uptake associated with inflammatory and granulomatous disease. To help establish the pathologic correlate of minimum ADC, we undertook a preliminary study investigating the relationship between minimum ADC and FDOPA PET uptake in patients with newly diagnosed GBM. Recent studies have shown that FDOPA has similar tumor uptake compared with methyl-C-L-methionine, resulting in a significant improvement in diagnostic accuracy compared with FDG-PET in evaluating both lowand highgrade tumors. Furthermore the kinetic modeling of FDOPA uptake has been shown to be useful for establishing tumor grade and has recently been shown to correlate with tumor Ki-67 proliferation indices in newly diagnosed gliomas. Our underlying hypothesis is that if minimum ADC corresponds to regions of high tumor cellularity, then tumoral regions of low ADC should overlap considerably areas of high FDOPA uptake. MATERIALS AND METHODS The institutional review board approved the study and written informed consent was obtained from each participant. Patients Data from 15 patients (10 men; age range, 47– 85 years) with histopathologically confirmed high-grade brain tumor (WHO grade III or IV) were used in this study. These patients were enrolled in a larger study aimed at developing FDOPA PET–MR imaging fusion-guided therapy for patients with primary brain tumors. Imaging Protocols Both the MR imaging and FDOPA PET studies were performed within 48 hours before tumor resection. MR imaging scans were performed by using a 3T Tim Trio scanner (Siemens, Erlangen, Germany). Routine diagnostic scans were supplemented with standard T1-weighted MR imaging scans (1-mm isotropic resolution) acquired before and after administration of contrast agent (0.1 mmol/kg of body weight, gadodiamide; Amersham Health, Oslo, Norway). Diffusion images were acquired by using highangular-resolution diffusion imaging with the following parameters: 60 axial sections; FOV, 30 30 cm; TR/TE, 9200/112 ms; 2.5-mm section thickness; acquisition matrix, 128 128 with a 2.3 mm in-plane image resolution; acceleration factor of 2; and maximum diffusion-encoding gradient strength of b 3000 s/mm . Sixty-five diffusion-weighted images were acquired at each location consisting of 1 low(b 0) and 64 high-diffusionweighted images. The acquisition time for the diffusion dataset was 9.67 minutes. A field map for the diffusion data was acquired by using two 3D gradient recalled-echo images (TE1/TE2, 4.76/ 7.22 ms) to assist the correction for distortion due to susceptibility-induced inhomogeneity. This acquisition protocol was selected to enable the study of tumor infiltration along white matter tracts by using tractography-based analysis techniques. FDOPA preparation took place in a radiochemistry laboratory by using a previously reported synthesis. PET imaging was performed by using a Gemini GXL scanner (Philips, Best, the Netherlands). An FDOPA activity of 151 MBq on average was administered intravenously (range, 138 –164 MBq). A low-dose transmission CT scan was performed followed by a 75-minute list mode acquisition. The images were reconstructed by using ordered subset expectation maximization, with corrections for attenuation and scatter. The final volume had a matrix size of 128 128, consisting of 90 planes of 2 2 2 mm voxels. Diffusion Processing To reduce image distortions and artifacts generated from involuntary head motion and physiologic noise, we used the following diffusion MR imaging preprocessing pipeline that has been fully described elsewhere. In brief, raw diffusion-weighted images were corrected for subject motion by using the method described by Bai and Alexander. Thus, the diffusion tensor was calculated from the (uncorrected) diffusion data. With the diffusion tensor information, a synthetic image was generated for every volume of the diffusion series. Every volume of the raw diffusion data was then aligned with the corresponding synthetic volume by using a 6-df registration performed with Advanced Normalization Tools (http://picsl.upenn.edu/ANTS), with the appropriate adjustment of the b-matrix. Susceptibility distortions were corrected by using the field map with FUGUE and Phase Region Expanding Labeler for Unwrapping Discrete Estimates in raw image space, both part of fMRI of the Brain Software Library (FSL; http://www. fmrib.ox.ac.uk/fsl/), with signal-intensity correction. Motion artifacts were identified and replaced by using the detection and replacement of outliers before resampling method in conjunction with the registration method by Bai and Alexander described above. ADC maps were then generated by using standard tools 2 Rose AJNR ● ● 2013 www.ajnr.org found within the MRtrix package (http://www.brain. org.au/software). FDOPA PET Processing and ADC Image Fusion The FDOPA images were rigidly registered to the CE MR imaging by using a block matching approach with 6 df. Normalization of the FDOPA scans was then performed by using the standardized uptake value ratio method, whereby each voxel was divided by the mean uptake in the cerebellum, a region that shows only nonspecific tracer uptake. The cerebellum was manually defined on the FDOPA maps. To register the FDOPA and ADC maps, we rigidly aligned b 0 images acquired as part of the High Angular Resolution Diffusion Imaging sequence to the coregistered CE MR imaging, by using the FSL Linear Image Registration Tool and 6 df, using mutual information. Generation of Minimum ADC and Maximum FDOPA SUVR Regions There is no widely accepted method to delineate tumor margins on the basis of FDOPA threshold levels, though the use of the percentage of maximum SUVR based on the tumor background ratio has been suggested. Techniques based on the use of a single global threshold are not suitable for FDOPA images due to possible uptake within noncancerous tissue (striatum) close to the tumor and inhomogeneous tracer uptake within the whole tumor volume. In this study, an experienced nuclear medicine physician (P.T.) manually defined the tumor boundary by using the fused CE MR imaging and FDOPA maps. Regions of interest on the ADC maps, outside of the FDOPA-defined tumor volume, were manually placed to identify ADC measures in normal tissue. This placement was performed, when possible, on several serial axial sections containing no FDOPA uptake (ie, infiltrating tumor). Reference to coregistered CE MR imaging and T2-weighted scans was also considered to avoid including tissue exhibiting significant compression effects due to excessive tumor growth. The mean (SD) volume of normal tissue was 172 54 mL. Regions of minimum ADC within the FDOPA-defined tumor volume were generated by applying a threshold of 450 10 6 mm/s (ie, 3 SDs lower than the mean ADC values derived for normal parenchymal tissue). Regions of maximum FDOPA uptake within the tumor volume were defined by voxels with the 20% highest SUVR. Mean FDOPA SUVR indices were generated for the tumoral regions exhibiting minimum ADC. Mean ADC values within regions corresponding to maximum FDOPA SUVR were also