|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
© 2003 American Society for Clinical Oncology Early Detection of Response to Radiation Therapy in Patients With Brain Malignancies Using Conventional and High b-Value Diffusion-Weighted Magnetic Resonance Imaging
From the Advanced Technology Center, Neurosurgery Department, Oncology Institute, and Plastic Surgery Department, Sheba Medical Center, Tel-Hashomer, Sackler School of Medicine and the School of Physics and Astronomy, Tel-Aviv University, Ramat-Aviv, Israel; and Department of Radiology, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA. Address reprint requests to Yael Mardor, PhD, The Advanced Technology Center, Sheba Medical Center, Tel-Hashomer, Ramat-Gan 52621, Israel; email: yael{at}tauphy.tau.ac.il.
Purpose: To study the feasibility of using diffusion-weighted magnetic resonance imaging (DWMRI), which is sensitive to the diffusion of water molecules in tissues, for detection of early tumor response to radiation therapy; and to evaluate the additional information obtained from high DWMRI, which is more sensitive to low-mobility water molecules (such as intracellular or bound water), in increasing the sensitivity to response. Patients and Methods: Standard MRI and DWMRI were acquired before and at regular intervals after initiating radiation therapy for 10 malignant brain lesions in eight patients. Results: One week posttherapy, three of six responding lesions showed an increase in the conventional DWMRI parameters. Another three responding lesions showed no change. Four nonresponding lesions showed a decrease or no change. The early change in the diffusion parameters was enhanced by using high DWMRI. When high DWMRI was used, all responding lesions showed increase in the diffusion parameter and all nonresponding lesions showed no change or decrease. Response was determined by standard MRI 7 weeks posttherapy. The changes in the diffusion parameters measured 1 week after initiating treatment were correlated with later tumor response or no response (P < .006). This correlation was increased to P < .0006 when high DWMRI was used. Conclusion: The significant correlation between changes in diffusion parameters 1 week after initiating treatment and later tumor response or no response suggests the feasibility of using DWMRI for early, noninvasive prediction of tumor response. The ability to predict response may enable early termination of treatment in nonresponding patients, prevent additional toxicity, and allow for early changes in treatment.
DIFFUSION-WEIGHTED magnetic resonance imaging (DWMRI) allows for noninvasive characterization of biologic tissues on the basis of the diffusion of water molecules. The application of DWMRI for early detection of response to anticancer therapy is a new field.1–10 Most of the published studies, however, were performed using animal models and had limited capacity for observing and quantifying the low-mobility water population. This article presents the results of a feasibility study in which DWMRI (line-scan DWMRI protocol)11 was used to detect early changes at the cellular level in patients undergoing radiation therapy for metastatic or primary brain tumors. The goal of the study was to evaluate additional information provided by high-diffusion weighting and a multiexponential model of diffusion. The DWMRI characteristics of biologic tissues seem to be best described in terms of two main water populations: a low-mobility water population, restricted by macromolecules, intracellular components, and cellular membranes, and a high-mobility water population that is mostly free of these impediments. It has been shown in vitro that there is an order of magnitude difference between the diffusion of low-mobility/intracellular/bound and high-mobility/extracellular/free water molecules.12,13 In vivo, however, the classification of low-mobility water as intracellular water is less clear.14 Because DWMRI is sensitive to changes in diffusion and in the ratio between the high- and low-mobility water populations, it is anticipated that DWMRI will detect early changes in the morphology and physiology of tissues after antineoplastic treatments. These changes occur with increased permeability of cell membranes, cell swelling, and cell lysis, which induce changes in the intracellular and extracellular water populations. DW images are usually obtained by acquiring conventional T2-weighted images with the addition of diffusion weighting, which filters out the signal from high-mobility molecules. Furthermore, it is possible to acquire DWMRI at different diffusion-weighting values. At low diffusion-weighting values, most of the signal from the tissue is present in the images, whereas only the signal from the fast-diffusing water molecules is filtered out. At high diffusion-weighting values, most of the signal is filtered out, and the signal remaining in the image mostly originates from the molecules with the lowest mobility. Conventional DWMRI is acquired with diffusion-weighting values roughly filtering out most of the signal from the extracellular water molecules. Conventional DWMRI enables the calculation of one apparent diffusion coefficient, ADCconvntl, which is the measured value of water molecules’ mobility in the tissue. This parameter is affected by both the low- and high-mobility water populations. By monitoring ADCconvntl alone, we were able in some cases to detect changes as early as 1 week after treatment initiation. The differentiation between the high- and the low-mobility water populations is obtained by acquiring DWMRI over an extended range of diffusion-weighting values. This allows for the calculation of the ADC and volume fractions of the two populations separately. Effective antitumor therapy is expected to cause cellular changes in the tumor consistent with increased membrane permeability in some cases and cell lysis in others. We anticipate that cell lysis will cause the low-mobility volume fraction, Vlow, to decrease because of a smaller intracellular volume fraction. Vlow may also decrease because of less bonding of the extracellular water to the surface of cell membranes. The ADC of the high-mobility population, ADChigh, is expected to increase because of the decreased extracellular restriction that results from the lower tissue density. Because the calculated variables are influenced by effects of exchange between the two water populations, we may also notice an increase in ADChigh and a decrease in Vlow as a result of higher permeability of cell membranes. A single diffusion parameter, ADCconvntl, derived from conventional DWMRI, might not be sensitive enough to detect the decrease in the low-mobility volume fraction. The additional information obtained by performing the biexponential fit on the high-diffusion data, Vlow, together with the separation between the two additive effects, the volume effect and the diffusion effect, may therefore make possible higher sensitivity to early changes in response to therapy. Because we anticipated that ADChigh should increase as a response to treatment and Vlow should decrease, we define a diffusion index (R) that is calculated from the high DW data:
The sensitivity of DWMRI to detect posttherapy changes in the water populations was enhanced by monitoring changes in R. The early changes in this diffusion index correlated significantly with response or lack of response to therapy as measured by standard MRI performed 7 weeks after initiation of treatment. These results suggest that DWMRI could predict response as early as 1 week after initiation of treatment, as opposed to standard MRI, which would show changes only 1 to 2 months later. Early knowledge of response to therapy is clinically useful and possibly will enable decisions to change treatment early in the course of therapy, thereby preventing unnecessary toxicity or prolonged ineffective therapy in nonresponding patients.
Patients and Treatment Six patients with a total of eight brain metastases (three melanoma, three lung, and two breast cancer metastases), one patient with an acoustic neuroma, and one with a glioblastoma multiforme were included in this study. All patients underwent scans once before treatment and at regular intervals thereafter. Six patients received single-fraction radiosurgery of 16 Gy. The remaining two patients, one with brain metastases from breast cancer and one with a glioblastoma multiforme, were treated with fractionated radiotherapy in daily doses of 2 Gy to a total of 40 or 60 Gy delivered over 4 or 6 weeks, respectively.
MRI System
Computers and Software
DWMRI Method
where, I and I0 are the signal intensities in the presence and absence of diffusion weighting, respectively; the strength of filtering is represented by the parameter b, the diffusion-weighting factor, which is expressed in seconds per square millimeter. Conventional DWMRI is acquired at b values on the order of 1,000 sec/mm2, which can be roughly described as filtering out mainly the signal from the extracellular or high-mobility water molecules. Assuming a single water population, it is possible to use conventional DWMRI to calculate (equation 2 For a two-population system, with high and low mobility, the attenuation of the water signal owing to diffusion weighting is given by:
where ADChigh and ADClow are the ADCs of the two water populations, and Vhigh and Vlow are their respective apparent volume fractions. The differentiation between the two water populations is obtained by acquiring a series of DWMRIs at different diffusion-weighting values, up to high values, such as b = 4,000 sec/mm2. The data are then analyzed and the ADCs of the two populations, as well as the apparent volume fractions, are calculated.16–20
Imaging Parameters The duration of a conventional DWMRI measurement at b = 5 sec/mm2 and at b = 1,000 sec/mm2 is 20 seconds per average per slice. In this study we used two averages; therefore, the scan lasted 40 seconds per slice. The number of slices varied from patient to patient and was chosen in a manner that covered the entire tumor, with an extra slice in each direction. The duration of a 14 b-value scan is 3 minutes and 47 seconds per slice. To save scan time, only three to four slices chosen in the center of the tumor were scanned. Therefore, the overall diffusion scan time was approximately 20 minutes. In addition, pre- and postcontrast T1-weighted images were acquired, as well as T2-weighted images. Thus the overall scan time of an average examination reached 40 to 50 minutes. In normal white matter, the diffusion of the water molecules is anisotropic and data must be acquired in at least three orthogonal directions and then averaged to obtain isotropic diffusion coefficients. In this work, data were acquired using a monodirectional diffusion scheme (described in detail in Gudbjartsson et al11) with all three gradients turned on at the same time because of the long acquisition times at 0.5 T. Because of the natural isotropy of cancer tumors and because head orientation was reasonably reproducible, relative ADC changes with time are measurable even though individual ADC values may be affected by anisotropy.
Data Analysis
Example of Early Detection of Response to Treatment Seen With Conventional DWMRI Figure 2  (top) shows conventional DWMR images (b = 1,000 sec/mm2) of a lung cancer patient with a brain metastasis treated with a single fraction (16 Gy) of stereotactic irradiation. The first DW image shows that the metastasis is a little darker than the rest of the brain, probably because of the necrotic nature of the metastasis, representing a higher content of free or high-mobility water molecules than in normal brain tissue. One day posttreatment the signal intensity in the treated area decreased (ADCconvntl increased), and this effect became more apparent with time. The posttreatment ADCconvntl normalized to the pretreatment value is plotted for the treated area as a function of time at the bottom of the figure. It can be seen that ADCconvntl increased after treatment. The 40% increase in ADCconvntl measured 7 days posttreatment was followed by an 89% decrease in tumor volume as determined from the contrast-enhanced T1-weighted images, acquired 55 days after treatment.
Example of Early Detection of Response to Treatment Requiring High Diffusion-Weighting Values Figure 3 shows an example of a patient with breast cancer metastatic to the brain who received 4 weeks of fractionated radiation therapy (40 Gy) to the entire brain. In this case, conventional DWMR images (acquired at b = 1,000 sec/mm2) did not show any clear changes indicating whether the patient responded to treatment. There was no significant change in the normalized ADCconvntl values (Fig 3, bottom).
To increase the sensitivity of DWMRI for early changes in tissue morphology, we acquired images with higher diffusion weighting (up to b = 4,000 sec/mm2) and calculated the parameters ADChigh, Vlow (equation 3  ), and R (equation 1).
The normalized values of ADCconvntl, R, and tumor volume relative to the pretreatment values are shown as a function of time after initiation of treatment at the bottom of Fig 3
Parameters of Lesions
Correlation Between Changes in Diffusion Parameters Measured 1 Week After Initiation of Treatment and the Change in Tumor Size Measured by Conventional MRI 7 Weeks Later The correlation between ADCconvntl normalized to the pretreatment values (measured on average 8 days after initiation of treatment) and later tumor response is presented in Fig 4 for all 10 lesions. The response of the tumor is determined from the ratio between the tumor volume (as measured on average 48 days after initiation of treatment) and the pretreatment volume, calculated from the contrast-enhanced T1-weighted images. The correlation is considered statistically significant (Pearson correlation, P < .006). The slope of the linear correlation function is -0.32. Figure 5 shows the same correlation for R (equation 1). The latter correlation is considered extremely significant (P < .0006) and the slope of the linear correlation function is -0.83.
Table 2 is a different presentation of the data shown in Figs 4 and 5. The response group consisted of tumors that decreased in volume 1.5 months posttherapy, and the no-response group consisted of tumors with increased or unchanged volumes. The mean normalized value of the diffusion parameter relative to the pretreatment value, measured an average of 8 days after initiation of treatment, is presented. The one-tailed P value of the unpaired nonparametric Mann-Whitney U test was .06 for ADCconvntl and .005 for R.
Fitting Considerations In most cases, the  2 value for a biexponential fit of the DWMRI data was smaller than that of a monoexponential fit, supporting the two-population (low and high mobility) assumption. In some cases, because of partial necrosis of the tumor, the signal-to-noise ratio became quite low at high b values (Fig 1). In these cases, the maximal b value used for the fits was less than 4,000 sec/mm2 (the maximal b value used for each lesion is listed in Table 1). In one case (the first lesion in the table), the pretreatment tumor was highly necrotic, causing the 2 value of the monoexponential fit to be smaller than that of the biexponential fit; thus, for this patient, only ADCconvntl was calculated from the b = 5 and 1,000 sec/mm2 data.
Recent studies with animal brain tumor models1–3 and a clinical study describing two brain tumor patients9 suggested that DWMRI could detect the development of necrosis, occurring after successful treatment, as shown by an increase in ADCconvntl calculated from the conventional DWMRI.
In our study, using similar methodology, we found a statistically significant correlation (P < .006, Fig 4
By extending the range of diffusion-weighting values, apparent diffusion coefficients ADChigh and ADClow and the volume fraction of the low mobility population can be obtained. These ADCs are closer than ADCconvntl to the actual diffusion coefficients, although they are still influenced by effects of exchange between the two water populations. This additional information, reflected in the index R, enabled us to achieve a higher sensitivity for response to treatment at an early stage. In the case presented in Fig 3 The index R produced a stronger correlation with tumor response than did the conventional parameter ADCconvntl. Moreover, the difference between the response versus no response groups was not significant for ADCconvntl, whereas it was for R. Finally, the linear correlation function for R was 2.5 times steeper than that for ADCconvntl. Using this approach, all six lesions, which later showed response by conventional MRI, showed early changes in the normalized value of the index R. For most of the lesions, the relative change in R was greater than the relative change in ADCconvntl, demonstrating the feasibility to obtain enhanced sensitivity to detect early changes after treatment using a range of high DW data. Our research was performed using a relatively low magnetic field of 0.5 T, relatively weak gradients of 10 mT/m, and the line-scan DW image sequence, resulting in a marginal signal-to-noise ratio in the DW images, especially in the high DW data. This led to larger errors in the fitted results. We regard this study as an initial demonstration of the feasibility of using high DW data to enhance the sensitivity to early cellular changes in response to treatment. In the future, we aim to test the feasibility of applying this method for clinical use. To obtain higher signal-to-noise ratios, we plan to work with higher external magnetic fields, which will increase the measured signal, and stronger gradient amplitudes, which in turn should allow for shorter diffusion gradient duration and diffusion times. This will result in less loss of signal owing to T2 relaxation effects and decrease of the effects of exchange between the two water populations.24 The data presented in this study were acquired using the line-scan DW image sequence. Although this approach is less susceptible to motion and susceptibility artifacts, it yields relatively poor signal-to-noise data per unit scan time and is not widely available. The methodology presented is generic in nature and could also be generated by any other DWMR sequence. For example, echo-planar imaging sequences, which are fast MRI sequences that are essentially immune to undesired motions, are widely used and are available on most new MR systems. This type of sequence would be most appropriate for a routine clinical use because of its availability and the short acquisition times. This study confirms that DWMRI, at the conventional diffusion-weighting value of b = 1,000 sec/mm2, can provide early information on the effects of radiation treatment not attainable by other conventional noninvasive means in patients with brain malignancies. We further demonstrated that the additional information obtained from data acquired up to high diffusion-weighting values of b = 4,000 sec/mm2 may enhance significantly the sensitivity for early detection of response to treatment. Most importantly, we have shown clinically that there is a significant correlation between early water diffusion relative changes measured 1 week after initiation of treatment and relative changes in tumor volume measured by conventional means 7 weeks later. This work demonstrates the feasibility of DWMRI to noninvasively detect changes in brain tumors as early as 1 week after initiation of radiation treatment. Moreover, these results indicate that high diffusion weighting and a multiexponential model of diffusion may provide additional information, enabling higher sensitivity to response to therapy. Early prediction of response or lack of response could allow early discontinuation of treatment, thereby reducing significantly unnecessary toxicity in nonresponding patients and allowing a more appropriate treatment to begin at an earlier stage.
DWMRI Method Application of a pair of pulsed magnetic field gradients sensitizes MR experiments to molecular diffusion or motion.15 In this method, the normalized intensity of the water signal is given by
where I and I0 are the signal intensities in the presence and absence of diffusion-sensitizing gradients, respectively,
By varying g,
In this case, ADChigh and ADClow are the ADCs of the high- and low-mobility water populations, and Vlow (where Vhigh + Vlow = 1) is the apparent volume fraction of the low-mobility water population.16–20 These parameters can only be calculated using DWMR images acquired up to high b values (in this case, up to b = 4,000 sec/mm2).
Supported by the Israel Science Foundation, the Israel Cancer Research Fund, Adams Super Center for Brain Studies at Tel-Aviv University, the Izmel program of the Israel Ministry of Industry and Commerce, and National Institutes of Health grant no. R01 NS39335.
1. Huang W, Roche P, Tudorica A, et al: The use of 1H MRS, diffusion and perfusion MRI in the evaluation of brain tumor response to therapy: Correlation between choline level and ADC, rCBV. Proc Int Soc Mag Reson Med 1:395, 2000 (abstr 395) 2. Chenevert TL, McKeever PE, Ross BD: Monitoring of early response of experimental brain tumors to therapy using diffusion MRI. Clin Cancer Res 3:1457–1466, 1997[Abstract] 3. Stegman LD, Rehemtulla A, Hamstra DA, et al: Diffusion MRI detects early events in the response of a glioma model to the yeast cytosine deaminase gene therapy strategy. Gene Ther 7:1005–1010, 2000[CrossRef][Medline] 4. Galons JP, Altbach MI, Paine-Murrieta GD, et al: Early increases in breast tumor xenograft water mobility in response to paclitaxel therapy detected by non-invasive diffusion magnetic resonance imaging. Neoplasia 1:113–117, 1999[CrossRef][Medline]
5. Chinnaiyan AM, Prasad U, Shankar S, et al: Combined effect of tumor necrosis factor related apoptosis inducing ligand and ionizing radiation in breast cancer therapy. Proc Natl Acad Sci U S A 97:1754–1759, 2000 6. Gupta RK, Sinha U, Cloughesy TF, et al: Inverse correlation between choline MRS signal intensity and the ADC in human glioma. Magn Reson Med 41:2–7, 1999[CrossRef][Medline] 7. Geschwind JF, Artemov D, Abraham S, et al: Chemoembolization of liver tumor in a rabbit model: Assessment of tumor cell death with diffusion-weighted MR imaging and histologic analysis. J Vasc Interv Radiol 11:1245–1255, 2000[Medline] 8. Jennings D, Hatton BN, Guo J, et al: Early response of prostate carcinoma xenografts to docetaxel chemotherapy monitored with diffusion MRI. Neoplasia 4:255–262, 2002[CrossRef][Medline]
9. Chenevert TL, Stegman LD, Taylor JM, et al: Diffusion magnetic resonance imaging: An early surrogate marker of therapeutic efficacy in brain tumors. J Natl Cancer Inst 92:2029–2036, 2000
10. Mardor Y, Roth Y, Lidar Z, et al: Monitoring response to convection-enhanced Taxol delivery in brain tumor patients using diffusion-weighted magnetic resonance imaging. Cancer Res 61:4971–4973, 2001 11. Gudbjartsson H, Maier SE, Mulkern RV, et al: Line scan diffusion imaging. Magn Reson Med 36:509–519, 1996[Medline]
12. van Zijl PCM, Moonen CTW, Faustino PJ, et al: Complete separation of intracellular and extracellular information in NMR spectra by diffusion weighting. Proc Natl Acad Sci U S A 88:3228–3232, 1991 13. Pilatus U, Shim H, Artemov D, et al: Intracellular volume and apparent diffusion constants of perfused cancer cell cultures, as measured by NMR. Magn Reson Med 37:825–832, 1997[Medline] 14. Thelwall PE, Blackband SJ: Ghosts, phantoms and spectra: Water diffusion and cell swelling in diffusion weighted imaging. Proc Int Soc Mag Reson Med 1:351, 2001 (abstr 351) 15. Stejskal EO, Tanner JE: Spin diffusion: Spin echoes in the presence of a time-dependent field gradient. J Chem Phys 42:288–292, 1965[CrossRef] 16. Niendorf T, Dijkhuizen RM, Norris DG, et al: Bi-exponential diffusion attenuation in various states of brain tissue: Implications for diffusion-weighted imaging. Magn Reson Med 36:847–857, 1996[Medline] 17. Mulkern RV, Gudbjartsson H, Westin CF, et al: Multi-component apparent diffusion coefficients in human brain. NMR Biomed 12:51–62, 1999[CrossRef][Medline] 18. Mulkern RV, Zengingonul HP, Robertson RL, et al: Multi-component apparent diffusion coefficients in human brain: Relationship to spin-lattice relaxation. Magn Reson Med 44:292–300, 2000[CrossRef][Medline]
19. Maier SE, Bogner P, Bajzik G, et al: Normal brain and brain tumor: Multicomponent ADC line scan imaging. Radiology 219:842–849, 2001 20. Mattiello J, Basser PJ, Le Bihan D: The b matrix in diffusion tensor echo-planar imaging. Magn Reson Med 37:292–300, 1997[Medline] 21. CN/ASD Group: MINUIT: Users Guide. Program Library, D506, CERN, Geneva, Switzerland, 1993
22. Le Bihan D, Breton E, Lallemand D, et al: Separation of diffusion and perfusion in intravoxel incoherent motion MR imaging. Radiology 168:497–505, 1988 23. Conturo TE, McKinstry RC, Aronovitz JA, et al: Diffusion MRI: Precision, accuracy and flow effects. NMR Biomed 8:307–332, 1995[Medline] 24. Karger J, Pfeifer H, Heink W: Principles and application of self-diffusion measurements by nuclear magnetic resonance. Adv Magn Res 12:1–89, 1988 Submitted May 10, 2002; accepted December 11, 2002.
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2003 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
= 31 ms, 
= 51 ms, TR = 2,907 ms, and TE = 105.2 ms. Diffusion curves were calculated from additional DWMR images obtained with 14 b values ranging from 15 to 4,000 sec/mm2, 
) signal intensity of viable tumor acquired at 14 b values up to b = 4,000 sec/mm2. The line is a biexponential fit. (
) necrotic tumor. The line is a monoexponential fit. (
) noise level.
is the gyromagnetic ratio of the nuclei, and g and 
