Monitoring Early Response of Experimental Brain Tumors to Therapy Using Diffusion Magnetic Resonance Imaging1

Quantitative magnetic resonance imaging was performed to evaluate water diffusion and relaxation times, Ti and T2, as potential therapeutic response indicators for brain tumors using the intracranial 9L brain tumor model. Measurements were localized to a column that intersected tumor and contralateral brain and were repeated at 2-day intervals before and following a single injection of i,3-bis(2cffloroethyl)-i-thtrosourea (i3.3 mg/kg). Tumor growth was measured using T2-weighted magnetic resonance imaging to determine the volumetric tumor doubling time (Td) before (Td = 64 ± i3h,mean ± SD,n = 16)andafter(Td = 75 ± 9 h, n = 4) treatment during exponential regrowth. Apparent diffusion coefficient of untreated tumors was independent of tumor volume or growth time, whereas relaxation times increased during early tumor growth. Diffusion displayed the strongest treatment effect and increased before tumor regression by 55% 6-8 days following treatment. Changes in relaxation times were also significant with increases of 16% for Ti and 27% for T2. Diffusion and relaxation times returned to pretreatment levels by i2 days after treatment. Histological examination supports the model that the observed increase in diffusion reflects an increase of extracellular space following treatment. Furthermore, the subsequent apparent diffusion coefficient decrease is a result of viable tumor cells that repopulate this space at a rate dependent on the surviving tumor cell fraction and recurrent tumor doubling time. Serial tumor volume measurements allowed determination of log cell kill of i.O ± 0.3 (n = 4). These results suggest that diffusion measurements are sensitive to therapy-induced changes in cellular structure and may provide an early noninvasive indicator of treatment efficacy. Received 3/3/97; revised 6/27/97; accepted 7/14/97. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. I This research was supported in part by Grant BE-149 from the American Cancer Society and NIH Grant R29 CA59009. 2 To whom requests for reprints should be addressed, at Department of Radiology-MRI, University of Michigan Hospitals, 1500 East Medical Center Drive, Ann Arbor, MI 48109-0030. Phone: (313) 936-8866; Fax: (313) 764-2412; E-mail: [email protected]. Introduction The management of high-grade malignant tumors of the central nervous system is problematic, and despite the use of multimodality therapy, malignant gliomas remain uniformly fatal (1). Furthermore, quantitation of the response of brain tumors to therapy is more difficult than in systemic cancers because an improvement in patient function is multifactorial, and changes in neurological deficits may be unrelated to changes in tumor volume (2-5). MRI3 and X-ray CT scans of malignant human brain tumors do not readily allow quantitation of the actual tumor volume. Administration of a contrast agent allows estimates of tumor areas from the largest cross-sectional area of contrast enhancement indicating a compromised bloodbrain barrier. The blood-brain barrier, however, may be altered by chemotherapeutic agents (5) or corticosteroids; thus, measurements of tumor dimensions by the boundaries of contrast enhancement is an indirect estimate of tumor size. Moreover, not all tumors exhibit contrast enhancement (6). Because of the difficulty in obtaining accurate measurements of intracranial tumor volumes, brain tumor therapy trials usually report median survival time and median time to progression as a quantitative measure of response (2-5). Improved methods that would allow for earlier and accurate quantitation of the therapeutic response in individual patients are still needed. Because diffusion MRI is sensitive to tissue structure at the cellular level, we believe that this technique has the potential to detect important quantitative information about the tumor cellular changes that would occur following successful therapeutic intervention. Cellularity and the integrity of cellular membranes that impede water translational mobility can affect the diffusion of water within tumor tissue. For example, water diffusion measurements have been shown to be sensitive to tissue cellular size, extracellular volume, and membrane permeability (7-9). In addition, it has been shown that structural anisotropy is manifest as diffusion anisotropy (10-12). Diffusion MRI applications indude imaging of cerebral stroke (13-16) and as a probe in the study of central nervous system tumors in humans (17-19). Diffusion studies on human (17-20) and animal (21-23) tumors have demonstrated strong differences between solid tumors relative to high diffusion within necrosis and cysts. As both solid tumor and peritumor edema can have a continuum of diffusion values that are elevated relative to normal brain, the diffusion coefficient alone does not readily distinguish tumor from edematous tissue (19, 23). However, structurally anisotropic tissues, such as white matter, can exhibit persistent diffusion anisotropy, 3 The abbreviations used are: MRI, magnetic resonance imaging; MR. magnetic resonance; CT, computed tomography; ADC, apparent diffusion coefficient; NMR, nuclear magnetic resonance; BCNU, 1,3-bis(2chloroethyl)-l-nitrosourea; ROl, region(s) of interest; IR, inversion recovery. Research. on November 11, 2017. © 1997 American Association for Cancer clincancerres.aacrjournals.org Downloaded from despite a diffusion coefficient elevated by edema. This could every 2 days. A rigid bite-bar was used to hold the anesthetized potentially provide contrast between edema and an isotropic rat within a 35-mm diameter transmit-receive slotted-resonator solid tumor ( 18-2 1 ). radiofrequency coil (28). Gradient-recalled-echo MRI interThe consistent observation of high diffusion in necrotic leaved in orthogonal planes was used to position the animal tissue relative to solid tumor provides a rationale for the use of rapidly and reproducibly. Quantitative in vivo measurements diffusion to monitor cellular changes following anticancer therincluded tumor volume, localized water diffusion, and localized apies. Preliminary studies in gliomas revealed that the evolution Tl and T2 relaxation times. The 9L glioma appears hyperintense toward necrosis that occurs following successful therapy could on T2-weighted MRI and has relatively distinct tumor margins be detected as an increase in ADC relative to pretreatment levels with only moderate pentumoral edema. Therefore, tumor volin the intracranial 9L model after administration of a chemoume was assessed using multislice T2-weighted MRI acquired therapeutic agent (24, 25). The observed increase in water with repetition time TR = 2850 ms, echo time TE = 80 ms, 32 mobility was anticipated to be the result of increased interstitial coronal 0.8-mm slices, 128 X 128 matrix, two-signal average, water volume and cellular permeability as the tumor cells neand a 30-mm field of view. These images were acquired as two crosed, thus indicating the potential of diffusion MRI for preinterleaved 16-slice sets collected near the beginning and at the dicting therapeutic efficacy. More recently, others have made end of the NMR session. This indicated whether the animal had similar observations in a s.c. tumor model using diffusionmoved due to failed sedation, in which case only the initial weighted spectroscopy (26). 16-slice set was used for tumor volume estimation. Tumor In this study, we sought to determine if quantitative MR volume was quantified as the product of slice-to-slice separation diffusion measurements could be used as a predictor of treatand the sum of areas from manually drawn tumor ROI on ment outcome through early detection of cellular changes before images. tumor regression in an intracerebral tumor model. In addition, Volumetric tumor doubling time, Td, and constant, K, reNMR relaxation times, Tl and T2, have been shown to be lated to initial viable tumor volume, were calculated using a sensitive to extracellular volume (23, 27) and thus may also two-parameter, least-squares-fit of measured tumor volume verdemonstrate changes in response to treatment. In this regard, sus time by: diffusion and Tl and T2 relaxation times were obtained from rat intracerebral 9L gliosarcomas before and following treatment Tumor volume (t) = K 2:/Td with BCNU. We found that of the MR parameters evaluated, diffusion MRI was the most sensitive to changes in membrane integrity affecting intraand extracellular water volumes following treatment. Furthermore, MRI diffusion values varied among x, y, and z directions, indicating the presence of diffusion Data during exponential pretreatment and tumor regrowth were fit independently to yield Tdpre and respectively. Pretreatment tumor volume, Vpre, and the effective volume of tumor that survives treatment, were derived by extrapolation of fitted pretreatment and regrowth curves to the time of treatment. anisotropy near tumor boundaries that correlated with histologj the cellular density of the tumor at time of treatment and ically distinct features. These results should help motivate future during exponential regrowth are approximately the same, then experiments correlating diffusion changes in human brain tumors with clinical outcome following therapeutic intervention. the following provides an estimate of log cell kill: Materials and Methods F 1 Log cell kill = log10l j I L postJ Glioma Model. Rat 9L tumor cells were grown as monoT2-weighted images were also used to prescribe subselayers in sterile plastic flasks in modified Eagle’s MEM with quent water diffusion and relaxation time measurements. The 10% fetal bovine serum. Cells were cultured in an incubator at diffusion pulse sequence used orthogonal 90#{176} and 1 80#{176} slice37#{176}C 95% air, 5% CO2 atmosphere until confluent. Cells were selective pulses that defined a right-to-left oriented, 2 X 2-mm then harvested by trypsinization, counted, and resuspended in column through the most homogeneous tumor region and conserum-free medium for intracerebral injection. tralateral brain ( 1 1 , 2 1). Diffusion measurement is susceptible to Adult male Fischer 344 rats weighing approximately I 50 g errors due to tissue motion. These artifacts were lessened by were anesthetized via i.p. injection of a ketamine (87 mg/kg using a frequency encode gradient along the column axis for body weight) and xlazine (13 mg/kg) cocktail. A small incision spatial encoding and magnitude processing of Fourier transover the right hemisphere was made, and a high-speed drill was formed echoes. As a result, phase cancellation effects comused to create a I -mm diameter burr hole through the skull. The monly associated with tissue motion in the presence of diffusion rat head was affixed in a stereotactic holder for inoculation of gradient pulses were greatly reduced (1 1). Spatial resolution l0 9L tumor cells in 5 p.! in the right forebrain at a depth of 3 along the one-dimensional image of the column was 0.23 mm. mm via a 25il Hamilton syringe. The surgical region was Potential directionality of water mobility is probed by the conirrigated with 70% ethanol, and the burr hole was filled with trolling diffusion gradient direction. To account for potential bone wax to reduce extracerebral extension of the tumor and diffusion anisotropy, paired diffusion sensitization gradient loss of cerebrospinal fluid. The skin incision was sutured closed, pulses were trapezoidal, 15 ms each, and applied independently and the rats were allowed to recover. Tumors were produced in on x, y, and z axes. Diffusion gradient direction and amplitude 20 rats for this experiment. were interleaved during the scan and stepped from 14 to 60 MRI Measurements. In vivo MRI was initiated 10-12 milliTesla/meter for a total of 42 “b-factors” (17) on each axis, days after tumor inoculation and performed serially on average ranging from 87 to 1 669 s/mm2. ADCs were calculated for all 1458 Therapy Assessment via Quantitative Diffusion MRI