| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Technical Developments |
1 From the Department of Radiology (J.P.D., D.M.P, L.H.S., D.B.), Weill Medical College of Cornell University, 1300 York Ave, Box 234, New York, NY 10021-4885; Departments of Radiology (D.M.P., L.H.S., J.A.K., D.B.), Surgery (J.H.H.), Pediatrics (P.A.M., R.G.), Pathology (A.G.H.), Biostatistics (H.T.T.), Medical Physics (J.A.K., D.B.), and Medicine (J.A.K.), Memorial Sloan-Kettering Cancer Center, New York, NY; and Institute of Neurology, University College London, England (P.S.T.). Received October 9, 2001; revision requested January 2, 2002; final revision received October 16; accepted November 6. Supported in part by National Institutes of Health grants R01HL50139, CA05826-038A1, and RO1CA62556. Address correspondence to J.P.D. (e-mail: jpd2001@med.cornell.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
© RSNA, 2003
Index terms: Bone neoplasms, MR, 41.12143, 44.12143, 45.12143 • Chemotherapy, 41.3221, 41.3281, 44.3221, 44.3281, 45.3221, 45.3281 • Ewing sarcoma, 41.3281, 44.3281, 45.3281 • Osteosarcoma, 41.3221, 44.3221, 45.3221
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
The ability of static clinical magnetic resonance (MR) images to depict tumor necrotic fraction during induction chemotherapy is limited by several factors. T1-weighted MR images alone cannot differentiate viable tumor from nonviable tissue or edema. T2-weighted MR images cannot adequately distinguish tumor from necrosis, and lesion boundaries are frequently overestimated because of the presence of edema and hemorrhage (4,5). It was determined that static T2-weighted MR images were an indicator of response following chemotherapy with an accuracy of 71% but a specificity of only 45.5% (6). Static T1-weighted contrast material–enhanced images can differentiate necrosis and hemorrhage from tumor but have difficulty separating tumor from chemotherapy-induced inflammation (7). Dynamic enhanced MR imaging has been shown to assist in detecting viable tumor in osteogenic sarcomas being treated with chemotherapy (8–10). Regions of necrosis, muscle, vessel, and viable tumor display distinct time-intensity curves in dynamic images. The rapid acquisition sequences used in these dynamic studies provide more information on vascular uptake than do static gadolinium–enhanced subtraction studies.
Advances in chemotherapy and surgical techniques currently allow 90% of adolescents with localized osteogenic sarcoma of the extremity to be treated with limb-salvage surgery instead of limb ablation (11). The relationship of a sarcoma to the joint, growth plate, and major neurovascular structures determines what tissue must be excised and what can be preserved. This influences the success of limb-salvage tumor surgery. Surgical margins are typically narrowest in these critical areas. Narrow margins, particularly in difficult tumor locations, such as the popliteal space, and insufficient tumor necrosis in response to chemotherapy contribute to the development of local recurrence (12–15). For example, a combination of good necrosis and wide surgical margins was associated with only one recurrence in 93 patients. In contrast, patients who had a poor necrosis response developed a local recurrence in four of 14 cases despite wide surgical margins (15). These data suggest that it may be important to tailor surgery on the basis of obtainable margins and the local response to chemotherapy. To avoid raising the risk of local tumor recurrence, clinicians need an accurate preoperative assessment of these factors. At present, the response of the tumor to chemotherapy is not reliably identified until after surgery. Therefore, the purpose of our study was to estimate tumor necrosis present in osteogenic and Ewing sarcomas being treated with induction chemotherapy before surgery by means of dynamic contrast-enhanced MR imaging.
|
Materials and Methods |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
To obtain the most accurate correlation with pathologically determined necrotic fraction, only patients who did not receive chemotherapy between the MR imaging examination and subsequent surgery were included. The resulting 29 patients (10 female and 19 male patients; age range, 6–35 years; mean age, 16 years ± 1 [standard error]) had osteogenic (n = 19) or Ewing (n = 10) sarcomas located in the distal femur or proximal tibia (n = 23), pelvis (n = 3), humerus (n = 2), and shoulder (n = 1).
Patient diagnosis was confirmed initially with a bone or soft-tissue biopsy. Patients with osteogenic sarcoma were treated with 8 weeks of high-dose methotrexate, cisplatin, and doxorubicin (2). Patients with Ewing sarcoma were treated with cyclophosphamide, doxorubicin, and vincristine. MR images were obtained in addition to thallium 201 and technetium 99m nuclear scans, chest CT scans, and radiographs of the tumor before en bloc resection of the entire tumor mass. Consolidation chemotherapy was given for 21 weeks after surgery. Postoperative chemotherapy was administered independent of the pathologically determined Huvos grade.
Chemotherapeutic protocols included written informed consent from all patients or their guardians and were approved by the institutional review board. Moreover, the board reviewed our manuscript and noted that our study would have been considered exempt research, if reviewed initially, and that patient informed consent was not required for this retrospective review.
MR Imaging
MR imaging examinations were performed with a 1.5-T MR system (Signa Horizon or LX; GE Medical Systems, Milwaukee, Wis). Patients underwent imaging before definitive surgery (15 days ± 2), at which time the tumor was resected for clinical reasons and the necrotic fraction was determined pathologically. A typical staging study was performed, including T1-weighted MR imaging (repetition time msec/echo time msec of 400/17) and fat-saturated T2-weighted MR imaging (4,000/85) in the transverse, sagittal, and coronal planes. The sagittal plane was prescribed for the dynamic enhanced portion of the study to view the feeding vessel all in one section. Initially, gadopentetate dimeglumine (Magnevist; Berlex Laboratories, Wayne, NJ) was administered manually into a peripheral arm vein, at a concentration of 0.1 mmol per kilogram of body weight, resulting in a standard dose of 1 mL of gadopentetate dimeglumine per 10 kg of patient weight. A 20-mL saline flush was administered at the same flow rate immediately after the gadopentetate dimeglumine bolus to flush the contrast material from the peripheral vein into the central venous system, resulting in faster and more uniform timing of the delivery of the contrast material bolus to the central venous system and subsequently to the arterial system (16). A power injector delivery system (Medrad, Indianola, Pa) was used subsequently to provide fixed flow rates. Patients with a central venous silicone catheter required a slower flow rate (0.8–1.0 mL/sec) to prevent damage to the catheter; in all other patients, contrast material was administered at a flow rate of 2 mL/sec. All patients received contrast material while they were at the imaging position in the bore of the magnet.
Dynamic perfusion MR images were acquired with a fast multiplanar spoiled gradient-echo sequence. The entire tumor was imaged contiguously with 10–12-mm-thick sections, yielding from five to nine sections (depending on tumor extent). Acquisition parameters included 9/2, 30° flip angle, 15.63-kHz receive bandwidth, 20–24-cm field of view, and 256 x 128 matrix (yielding a voxel resolution of 12–20 mm3). These parameters provided a temporal resolution between 4.75 and 9.45 seconds per image and approximately five points on the initial slope of the blood pool time-intensity curve. This temporal resolution was sufficient to observe the initial uptake of gadopentetate dimeglumine into the region (17,18). Data were acquired at a total of 20–40 time points in imaging times of less than 5 minutes. In all cases, a central section was placed parallel to and directly over the long axis of the bone. A preliminary imaging sequence performed at five time points ensured adequate signal-to-noise ratio and exact section positioning before administration of gadopentetate dimeglumine.
Pathologic Examination
The gross specimen was oriented in the coronal direction, and a radiograph was obtained to determine the section that demonstrated extension of the tumor into soft tissue. As part of the standard examination of these tumors, much of the soft tissue was removed, and a 5-µm-thick slice parallel to the first cut was removed for complete pathologic analysis. Approximately 20–60 tissue slides were taken from the single slice for determination of the necrotic fraction. The response grading of the tumor, as outlined in the Table, was performed or supervised by one pathologist (A.G.H.), with more than 30 years of experience, and used to estimate the amount of necrosis induced by chemotherapy. A clinically good response is considered to be at least 90% necrosis (ie, grade III or IV) (3). It should be noted that the occurrence of spontaneous necrosis within osteogenic sarcomas has been studied and rarely accounts for more than 25% of the total necrotic fraction (19); thus, any finding of substantially greater necrotic fraction in patients undergoing chemotherapy may be attributed to the effects of the therapy.
|
|
Analysis was performed for each voxel within a region of interest corresponding to the reconstructed coronal section placed over the central axis of the bone to compare with the pathology specimen. The region-of-interest boundary was defined primarily by one investigator (J.P.D.), but radiologists (D.M.P., L.H.S.) were consulted for clarification of the boundaries in some of the 29 studies if the margins of the mass were unclear; in those few cases, the tumor had largely resolved after therapy. In the majority of cases, boundaries of remaining tumor mass were evident. Regions of interest were selected to encompass as large a region as possible while avoiding partial volume effects and any substantial artifact; their size ranged from 2.2 to 91.3 cm2 for the 29 patients. A histogram containing the slope value for each voxel in the region of interest was created for the entire tumor, as well as for a single coronal section, with a bin value of 5% uptake per minute. All histograms were normalized to the number of voxels contained within each region of interest to account for variability in the size of tumor burden between patients.
A second method of analysis, a pharmacokinetic model, was also studied. The two-compartment model proposed by Hoffman et al (20), based on that of Brix et al (21), incorporates rate constants of gadopentetate dimeglumine transfer between the lesion and plasma compartments (kep) and elimination by the plasma (kel) (22). Assumptions in this model are that a monoexponential function describes the plasma concentration up to 20 minutes after injection; direct measurement of the plasma curve is not needed, as the clearance rate (kel) is estimated directly from the measured tissue curve (22). After a bolus injection of length 
, and assuming kep << 1 and kel << 1, the Hoffman initial equation reduces to Equation (2) (22). Additionally, assumptions about the in vivo tissue relaxivities and longitudinal relaxation times are not made but are contained in the fitted parameter (A) as shown in Equation (2). The model contains three fitted parameters: A (amplitude), kep (per minute), and kel (per minute):
|
At short time intervals after injection, the right side of Equation (2) is approximated by (1 + Akept); thus, the initial slope becomes proportional to Akep (18,22). This product was determined for each voxel in the region of interest and placed in a histogram.
The asymmetric continuous distribution described by Equation (3) is known as the extreme value (or Gumbel distribution) and was fitted to the histograms of both initial slope and compartmental model parameters (23).
|
The variable x represents the ordinate of the histogram containing bin values of percentage uptake per minute, or Akep. The variable y represents the abscissa of the histogram containing the fraction of voxels at each bin value. The amplitude (
), width (
), and median (µ) of the distribution were determined from the fitting procedure. The amplitude described the percentage of voxels within the tumor having the median value of initial slope or Akep, respectively. The width of the distribution was an indicator of the heterogeneity of the specific tumor. Tumors that were composed of a more uniform distribution of values presented a narrow width. The median value of the distribution characterized the value having the single highest concentration of voxels. Histograms were normalized to the total number of tumor voxels to allow direct comparison between patients. One justification for choosing this function was the heterogeneity of uptake present within the tumors. All tumors produced an asymmetric histogram with a tail caused by a component of greater uptake.
All data were fit to the corresponding model or histogram equations by using software developed in house with the interactive data language. A gradient-expansion algorithm within the language computes a nonlinear least-squares fit to a user-supplied function with an arbitrary number of parameters. Partial derivatives are calculated analytically and used in the fitting process to determine successive iterations until the 
2 value changes by a specified amount or until a maximum number of iterations have been performed.
The reproducibility and robustness of the compartmental model were examined to ensure that variations in the model parameters could be attributed to changes in the time-intensity curve and not to variations in signal-to-noise ratio levels. Randomly distributed noise was added to the time-intensity curves at 5% intervals up to 30%, which is expected to be the maximum observed in clinical images for this gradient-echo sequence. A set of compartmental fits (n = 100) were performed for each noise level for the Akep product.
Each patient was assigned a single MR imaging parameter of uptake slope and compartmental model (Akep) on the basis of the histogram analysis of all tumor voxels. These parameters were always compared with the pathologically determined necrotic fraction. Patients were placed into two groups: responders (≥90% necrosis) and nonresponders (<90% necrosis). The MR imaging methods of slope and compartmental model were analyzed independently with receiver operating characteristic curves to determine a cutoff for the MR imaging parameter that would correlate with the groups determined on the basis of necrotic fraction.
Statistical Methods
The overall degree of relationship between each parameter and percentage necrosis, as determined at pathologic examination, was assessed by means of the Pearson product moment correlation coefficient, R. The criterion for statistical significance was a two-sided test at a P value of less than .05. Correlations were also compared for each tumor type separately. Receiver operating characteristic curves were computed to display the sensitivity and specificity of a parameter in predicting response over their full range of possible cutoff values. The standard of reference for receiver operating characteristic analyses was the histologic necrosis determination. Receiver operating characteristic curves were computed to display the sensitivity and specificity of the MR imaging necrosis parameter in predicting response over the full range of possible cutoff values. The cutoff values were automatically calculated with a software program (SAS; SAS Institute, Cary, NC) routine written by one of the authors (H.T.T.). The area under each receiver operating characteristic curve of the slope and compartmental model methods was calculated. The cutoff values were then calculated to achieve the maximum percentage accuracy, as well as to maximize the lesser of sensitivity and specificity. Positive and negative predictive values were reported for optimal cutoff values of the receiver operating characteristic curves for both the initial slope and model analysis parameters.
|
Results |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
|
|
|
|
|
|
|
|
|
Discussion |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
A primary advantage in the model analysis is that the shape of the entire time-intensity curve is incorporated and not just the initial slope. Fitting of the entire time course allows curves with similar initial slopes but different rates of washout to be separated. An additional advantage of the model analysis is that it may be adjusted and expanded to accommodate new information that may be gained about the physiologic processes involved in gadopentetate dimeglumine tracer uptake within bone sarcomas. Compartmental modeling also affords the potential to assess physiologically important parameters that are directly related to vascular permeability, vascular perfusion, and extravascular volume (18). Limitations in the model approach are more complex analysis routines and the need for a deeper knowledge of tracer analysis and contrast agent mechanisms. Another difficulty in this model is the inability to determine whether changes in uptake in regions of necrosis are due to variances in flow or permeability. In addition, Tofts (22) has discussed the relevance of incorporating tumor T1 into the modeling scheme. This would allow decreases in slope and Akep to be attributed to endothelial permeability instead of T1. In the present study, although the slope and model methods provided similar correlation with pathologic findings, the inclusion of permeability and vascular volume quantitation may increase the accuracy of the model results. In our study, seeing the entire curve and using a compartmental model was not shown to have an advantage over using just the (simpler) initial slope, perhaps because the complete washout characteristics were not obtained (as a result of insufficient temporal resolution or to insufficient imaging duration).
Factors to be accounted for in understanding the physiologic and angiogenic bases for changes seen in dynamic enhanced MR imaging include, but are not limited to, microvessel density (or fractional blood volume), vascular permeability, and perfusion (25). Knopp et al (26) studied the pathophysiologic basis of dynamic enhanced MR imaging in breast tumors and found significant correlation with the compartmental modeling parameter kep and increased expression of vascular endothelial growth factor, a biologically active molecule. It was shown that the primary factor that affected contrast enhancement was a change in vascular permeability caused by increased leakage stimulated by expression of vascular endothelial growth factor. This suggests that dynamic enhanced MR imaging might also be valuable in the assessment of angiogenic changes as precursors to event-free survival; further study is needed.
The data correlation weakened because of inherent physical differences between the MR imaging method and the pathologic method, as well as the fact that dynamic enhanced MR imaging and pathologic examination measure different indicators of tumor necrosis (tissue enhancement and histologic tumor necrosis, respectively). MR imaging cannot depict areas of microscopic disease that were included in the necrotic fraction during pathologic examination. Partial volume effects due to the differences between the 10–12-mm MR imaging section and the 5-µm slice taken during pathologic examination likely contributed to the error. The ability of MR imaging to depict tissue throughout the entire heterogeneous tumor compared with the single slice provided at pathologic examination might also contribute to differences in the results. The sampling of a large tumor volume at pathologic examination is prohibitive but is routinely performed with MR imaging. In the present study, the total tumor necrotic fraction as determined at MR imaging was similar to that of a single-plane necrotic fraction, as determined at MR imaging or at pathologic examination.
The clinical utility of this method lies in the ability to noninvasively predict during therapy, in the large majority of cases, whether the tumor has responded to induction chemotherapy with at least 90% necrosis. Although the factors discussed earlier prevent a more accurate estimation of grade I or II responders to chemotherapy, that fact is overshadowed by the knowledge that event-free survival is significantly correlated only with responses of grade III and IV. It is hoped that additional advances in MR imaging technology, compartmental modeling, and other correlative indicators will provide an opportunity to improve the prognostic value of this study to further improve patient care.
In patients with osteogenic or Ewing sarcomas who are undergoing chemotherapy before surgery, the use of dynamic contrast-enhanced MR imaging for the estimation of percentage tumor necrosis may be a valuable method for assessment of treatment response and prognostication. The addition of parametric information relating to tumor uptake may provide information to the surgeon indicating possible extent of disease near specific surgical margins. Pathologic evaluation of necrosis following definitive surgery can be performed only once and is typically carried out only after a prolonged period of induction chemotherapy. Noninvasive and nondestructive testing of tumor viability can be performed early in the course of therapy and serially throughout therapy. This may allow earlier identification of inferior responders, giving clinicians the opportunity to identify a population of patients that may benefit from a change in therapy type or intensity. The goal of the present study was to eventually increase the chances for such interventions to improve subsequent patient outcome. We believe the ability of MR imaging to noninvasively depict the entire tumor to potentially demonstrate clinically relevant patterns of enhancement makes this method attractive for continued investigation.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  M. S. Kim, S.-Y. Lee, W. H. Cho, W. S. Song, J.-S. Koh, J. A. Lee, J. Y. Yoo, and D.-G. Jeon Tumor Necrosis Rate Adjusted by Tumor Volume Change Is a Better Predictor of Survival of Localized Osteosarcoma Patients Ann. Surg. Oncol., March 1, 2008; 15(3): 906 - 914. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Muruganandham, M. Lupu, J. P. Dyke, C. Matei, M. Linn, K. Packman, K. Kolinsky, B. Higgins, and J. A. Koutcher Preclinical evaluation of tumor microvascular response to a novel antiangiogenic/antitumor agent RO0281501 by dynamic contrast-enhanced MRI at 1.5 T. Mol. Cancer Ther., August 1, 2006; 5(8): 1950 - 1957. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. C. Thoeny, F. De Keyzer, V. Vandecaveye, F. Chen, X. Sun, H. Bosmans, R. Hermans, E. K. Verbeken, C. Boesch, G. Marchal, et al. Effect of Vascular Targeting Agent in Rat Tumor Model: Dynamic Contrast-enhanced versus Diffusion-weighted MR Imaging Radiology, November 1, 2005; 237(2): 492 - 499. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. van Vliet, C. F. van Dijke, P. A. Wielopolski, T. L. M. ten Hagen, J. F. Veenland, A. Preda, A. J. Loeve, A. M. M. Eggermont, and G. P. Krestin MR Angiography of Tumor-related Vasculature: From the Clinic to the Micro-environment RadioGraphics, October 1, 2005; 25(suppl_1): S85 - S97. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
