- Email: [email protected]
Correlation of PET standard uptake value and CT window-level thresholds for target delineation in CT-based radiation treatment planning
Int. J. Radiation Oncology Biol. Phys., Vol. 67, No. 3, pp. 720 –726, 2007 Copyright © 2007 Elsevier Inc. Printed in the USA. All rights reserved 0360-3016/07/$–see front matter
doi:10.1016/j.ijrobp.2006.09.039
CLINICAL INVESTIGATION
Lung
CORRELATION OF PET STANDARD UPTAKE VALUE AND CT WINDOW-LEVEL THRESHOLDS FOR TARGET DELINEATION IN CT-BASED RADIATION TREATMENT PLANNING ROBERT HONG, M.D.,* JAMES HALAMA, PH.D.,† DAVIDE BOVA, M.D.,† ANIL SETHI, PH.D.,* AND BAHMAN EMAMI, M.D.* *Department of Radiation Oncology, and †Department of Radiology, Nuclear Medicine Division, Loyola University Medical Center, Maywood, IL Purpose: To develop standardized correlates of [18F]fluoro-2-deoxy-D-glucose positron emission tomography (FDG-PET) standard uptake value (SUV) to computed tomography (CT)-based window and levels. Methods and Materials: Nineteen patients with non–small-cell lung cancer who underwent imaging with positron emission tomography (PET) and CT were selected. A method of standardizing SUV within CT planning software was developed. A scale factor, determined by a sensitivity calibration of the PET scanner, converts voxel counts to activity per gram in tissue, allowing SUVs to be correlated to CT window and levels. A method of limiting interobserver variations was devised to enhance “edges” of regions of interest based on SUV thresholds. The difference in gross tumor volumes (GTVs) based on CT, PET SUV > 2.5, and regions of 40% maximum SUV were analyzed. Results: The mean SUV was 9.3. Mean GTV volumes were 253 cc for CT, 221 cc for SUV > 2.5, and 97 cc for SUV40%Max. Average volume difference was ⴚ259% between >2.5 SUV and CT and ⴚ162% between SUV40%Max and CT. Percent difference between GTV > 2.5 SUV and SUV40%Max remained constant beyond SUV > 7. For SUVs 4 – 6, best correlation among SUV thresholds occurred at volumes near 90 cc. Mean percent change from GTVs contoured according to CT (GTV CT) was ⴚ260% for GTV2.5 and ⴚ162% for GTV40%Max. Using the SUV40%Max threshold resulted in a significant alteration of volume in 98% of patients, while the SUV2.5 threshold resulted in an alteration of volume in 58% of patients. Conclusions: Our method of correlating SUV to W/L thresholds permits accurate displaying of SUV in coregistered PET/CT studies. The optimal SUV thresholds to contour GTV depend on maximum tumor SUV and volume. Best correlation occurs with SUVs >6 and small volumes <100 cc. At SUVs >7, differences between the SUV threshold filters remain constant. Because of variability in volumes obtained by using SUV40%Max, we recommend using SUV > 2.5 for radiotherapy planning in non–small-cell lung cancer. © 2007 Elsevier Inc. Radiotherapy, Positron emission tomography (PET), Standard uptake value (SUV), Target delineation, NSCLC.
INTRODUCTION
modulated RT, stereotactic radiosurgery, and image-guided RT depend on precise tumor volume delineation, which is again dependent on radiologic imaging. Although computed tomography (CT) imaging techniques provide anatomic information in addition to electron densities of various tissues necessary for dose distribution information in computerized treatment planning systems, the advantage over other modalities is relative, especially in distinguishing malignancy from normal tissues in the context of atelectasis or pleural effusion. [18F]fluoro-2-deoxy-D-glucose positron emission tomography (FDG-PET) is a functional imaging modality that is increasingly being used in all aspects of oncology, including
Patients with unresectable or medically inoperable non– small-cell lung cancer (NSCLC) have a poor prognosis. These patients who receive radiotherapy (RT) with or without chemotherapy have a 5-year survival between 0% and 30% (1– 4). Although largely dependent on tumor stage and performance status, published literature suggests tumor control is improved with dose escalation using conformal radiation therapy (5–7). New treatment techniques are increasingly being used to allow additional dose escalation beyond those achieved by conventional three-dimensional conformal radiotherapy. Treatment modalities such as intensity-
Society for Therapeutic Radiology and Oncology (ASTRO), in Philadelphia, PA, November 5–9, 2006. Conflict of interest: none. Received Aug 14, 2006, and in revised form Sept 20, 2006. Accepted for publication Sept 21, 2006.
Reprint requests to: Bahman Emami, M.D., Loyola University Medical Center, Department of Radiation Oncology, 2160 S. 1st Avenue, Maguire Building, Room 2944, Maywood, IL 60153. Tel: (708) 216-2555; Fax: (708) 216-6076; E-mail: [email protected] Presented in part at the 48th Annual Meeting of the American 720
Correlation of PET standard uptake value and CT window-level thresholds
diagnostic evaluation of primary lung nodules, staging of mediastinal lymph nodes, and detection of distant metastases. In an overview of the available literature, FDG-PET was found to have a 91% sensitivity, 68% specificity in diagnosing primary lung cancer, a 83% sensitivity and 91% specificity in mediastinal staging (8), compared with 56 – 65% sensitivity and 73– 87% specificity for mediastinal staging for CT (9). MacManus et al. reported in an interim analysis of a prospective protocol evaluating radical radiotherapy for NSCLC, that nearly a third (5/18) of patients were excluded owing to the identification of metastatic disease or extensive intrathoracic disease by PET. Additionally, in half (6/12) of those proceeding to RT planning, ⬎25% of the PET/CT gross tumor volume (GTV) was not covered by the CT GTV (10, 11). Fused images of PET and CT studies have the cumulative benefit of providing physiologic data with precise topographic localization. Although there are reports of initial studies incorporating FDG-PET into RT treatment planning, there is no consensus on the use of standard uptake value (SUV) in target delineation (12–15). For lung nodules, in addition to size, shape, and location, an SUV ⬎2.5 is a commonly used threshold indicative of malignancy (16, 17). Others have used a preclinical procedure described by Erdi et al. that used phantom studies to accurately define lesions by setting the upper window level to the maximum intensity voxel and the lower level set at a value of 42% of the upper level for spherical lesions of ⬎4 cc (18, 19). Black et al. also used phantom studies to describe a method of defining GTV based on a linear regressive function threshold of SUV that resulted in smaller deviation when compared with a fixed image intensity threshold (42% of maximum intensity) (20). It is clear that with PET-defined tumor volumes in RT treatment planning, variation in setting image signal thresholds can significantly affect the contour of the GTV, resulting in considerable interobserver and intraobserver variation (21–23). Moreover, there is no reliable correlation of CT window and level settings to PET SUV such that PET metabolic data in the form of SUV can be consistently viewed in CT-based treatment planning software. A retrospective analysis investigating the impact of GTV delineation for NSCLC according to varying PET SUV thresholds is described. Fused image sets from separately acquired FDG-PET with CT simulation were used. This study provides a novel method of correlating CT window and level values to PET SUV values, for precise visualization of PET data. Furthermore, by modifying our visualization technique, we have developed a means to filter PET data to segment regions of interest based on user-defined SUV boundaries, effectively sharpening the tumor edges. The goal of this study is to describe a practical method for GTV delineation based on FDG-PET SUV visualization to aid radiation oncologists in image-based RT treatment planning.
●
R. HONG et al.
721
METHODS AND MATERIALS Study population This study was approved by Loyola University Medical Center’s Institutional Review Board. Each patient was required to have pathologic confirmation of NSCLC, locally advanced unresectable or medically inoperable with a Karnofsky performance status of 70 –100. Nineteen consecutive patients who underwent staging PET evaluation between December 2005 and June 2006, who also were planned for CT-based three-dimensional conformal radiotherapy, were included in the current analysis.
Patient simulation Patients were CT simulated supine with arms overhead with upper body custom alpha cradle immobilization (Smithers Medical Products, North Canton, OH). The alpha cradle was fabricated to fit within the PET scanner aperture, which required a smaller width and flat table top. Five CT and PET compatible fiducial markers were also placed in noncoplanar positions on the cradle to aid in coregistration of image sets in the axial, sagittal, and coronal planes. Fiducial markers consisted of radiopaque “donuts” which housed FDG-soaked gauze at the time of PET scan. An isocenter was localized for each patient. After a scout study, the patient was given 1 tablespoon of barium paste by mouth and 100 cc of intravenous contrast. The scan was performed immediately using a Philips AcQSim CT scanner (Philips Medical Systems N.A., Bothell, WA) acquiring images at 5 mm ⫻ 5 mm intervals (512 ⫻ 512 matrix, 0.94 mm pixel size) from the second cervical vertebral body to the liver over 45 s. The patients were scanned while breathing freely (Fig. 1). Patients then proceeded for PET scan with the compatible immobilization devices within 24 – 48 h of CT scan acquisition. The patients were scanned with a dedicated whole-body Philips
Fig. 1. (a) Patients were simulated with computed tomography (CT) and positron emission tomography (PET) within 72 h. Custom fabricated immobilization was the same for each scan. (b) Four noncoplanar fiducials were placed on the alpha cradle. At the time of positron emission tomography simulation, [18F]fluoro-2deoxy-D-glucose was embedded in the center, allowing for improved image registration.
722
I. J. Radiation Oncology
●
Biology
●
Physics
Volume 67, Number 3, 2007
Fig. 2. (a) Schematic representation of positron emission tomography standard uptake value and computed tomography window and level (voxel) relationship. (b) Narrowing the displayed window to filter [18F]fluoro-2-deoxy-D-glucose positron emission tomography image to allow enhanced visualization of volume edges.
Allegro System PET camera. All patients had their height and weight determined at the time of study. Four-hour resting glucose values were measured and determined to be within acceptable ranges (⬎65 to ⬍200). Intravenous access was obtained. FDG quantity was based on weight and ranged between 6 mCi and 14.5 mCi. With gentle intravenous fluid hydration, FDG was injected intravenously and isotope was distributed over 90 min in a resting state. The patient was positioned and fiducials were prepared with gauze-soaked FDG. Scan acquisition occurred over 45 min.
PET interpretation The PET study was qualitatively reviewed by an experienced nuclear radiologist in comparison with prior diagnostic chest CT studies. Generally, regions of increased and abnormal FDG uptake beyond expected normal tissue uptake were characterized as tumor when SUV ⬎ 2.5. However, this qualitative assessment took all clinical and imaging factors into consideration including size and location of the lesion, histologic differentiation of the tumor, and inflammatory processes. For example, a region of “streaking” SUV ⬎ 2.5 was detected in the supraclavicular fossa of a patient that was qualitatively interpreted as a benign inflammatory response. Likewise, a malignant-appearing 1-cm nodule evaluable on CT but with SUV ⬍ 2.5 was scored as tumor. This qualitative binary assessment of the PET study was performed in the presence of the treating radiation oncologist and considered in the final gross tumor target delineation. In PET, the raw data corresponds to a count of the number of positron annihilations in a voxel. A scan-specific scale factor, determined by a sensitivity calibration (SUVcal) of the PET raw data, is used to convert voxel counts to activity per gram in the tissue. The ratio between the SUVcal (i.e., raw data) and SUV for any voxel in the PET image is constant. The SUV is then the ratio of the activity per gram in the tumor to the administered activity per gram of total body mass (Fig. 2). Both the maximum tumor SUV and SUVcal for the PET study were recorded.
PET data were initially visualized using arbitrary window and level values defined subjectively to ensure proper image registration. Then, based on the maximum tumor SUV and the minimum threshold SUV, window and level values were generated based on the PET sensitivity calibration factor. The minimum SUV thresholds that were evaluated were regions ⬎2.5 and 40% maximum SUVs based on published literature recommendations (16 –19). Although this initial filter allowed qualitative visualization of SUV of the full range of activity defined by the minimum and maximum thresholds, the “edges” of the uptake were not clearly defined. By narrowing the displayed window values to within 0.1 of the minimum SUV, the regions at the periphery of the area of interest could be sharpened (Fig. 3). This allowed for precise contouring of the user-defined regions based on the available PET data (Fig. 4). The CT alone tumor volumes consisted of the primary tumor and any regional lymph nodes with a diameter of ⬎1 cm on the short axis. The window width and level in the CT images were 1,600 Hounsfield units (HU) and ⫺600 HU for the lung parenchyma and 400 HU and 20 HU for the mediastinum. Total tumor volume for the GTVs contoured using CT alone (GTV CT), PET SUV ⬎ 2.5 (GTV25), and PET SUV 40% maximum (GTV40) were generated using XIO planning software.
Statistical analysis The GTV volumes generated by the two SUV thresholds (GTV25, GTV40) were compared with each other as well as the
Target delineation Positron emission tomography Digital Imaging and Communications in Medicine (DICOM) data were sent to the radiotherapy planning software (CMS, Inc. Focal and XIO v4.3.1) where a combination of automated fusion using Focal’s Mutual Information algorithm and manual registration of overlapping fiducials was performed. Maximum tumor SUV values and the scale factor, determined by a scan-specific sensitivity calibration of the PET image, were used to convert voxel counts to SUV (Fig. 2). The
Fig. 3. Illustration of the variables and formulas used to generate window and level values to display standard uptake value (SUV) in the computed tomography– based radiotherapy planning software. (Available online at http://www.cuberick.com/⬃faisalv/ window-level-suv-correlator.html).
Correlation of PET standard uptake value and CT window-level thresholds
●
R. HONG et al.
723
Fig. 4. Example of gross tumor that is difficult to visualize on computed tomography with stepwise visualization of gross tumor volume based on coregistered positron emission tomography scan: (a) represents fused images displayed with a full range of standard uptake value (SUV); (b) represents fused images only displaying regions between SUV 2.5 and SUV max; (c) represents fused images with a narrowed window displaying SUV 2.5–2.6, resulting in sharpened edges for the regions of interest.
GTV CT. The percentage difference between GTV25 and GTV40 was calculated. The percentage difference between GTV25 and GTV CT as well as GTV40 and GTV CT was also calculated based on absolute volumes. Linear regression analysis, box-whisker plot visualization and graphs were generated by Systat (Version 11; Systat Software Inc., Point Richmond, CA).
RESULTS The mean SUV for the tumors examined was 9.3 (median, 9.0, range, 4.4 –14.8). The mean GTV volumes were 253 cc (range, 12–922 cc) on CT, 221 cc (range, 1–920 cc) for SUV ⱖ 2.5 and 97 cc (range, 4 –324 cc) for SUV 40%Max. The average volume difference was ⫺259% between the ⱖ2.5 SUV and CT GTV and ⫺162% between the SUV 40% Max and CT GTV. Overall, using a 25% volume difference as a cutoff for significant alteration of GTV CT, 58% (11/19) of the GTV25 and 95% (18/19) of GTV40 volumes were substantially different from the original CTbased tumor volumes. The two largest differences occurred in patients whose tumor was delineated within atelectatic lung parenchyma. An analysis between the two PET thresholding techniques was also performed. When the percentage difference between the tumor volumes is plotted against their respective maximum SUV (Fig. 5), the percent difference between GTV ⱖ 2.5 SUV and SUV 40% Max remains constant after SUV ⬎ 7, with the GTV 40% Max volume averaging ⫹ 55%. There is a clear plateau on the shape of this curve which denotes minimal difference in volume between the
two SUV thresholding methods at very high SUV values. Conversely, there is a steep slope denoting considerable variation in tumor volumes between the two techniques at nominal SUVs between 4 and 6. To further analyze whether there was any correlation with tumor volume, the GTV volumes as determined by the two thresholds were plotted against the maximum SUV for each respective patient. Using a linear regression analysis, the best correlation of GTV among the SUV threshold values occurs at small tumor volumes near 90 –100 cc for SUVs between 4 – 6 (Fig. 6).
Fig. 5. Relationship of the percentage difference between the gross tumor volume (GTV) contoured based on a standard uptake value (SUV) ⬎2.5 vs. SUV 40% Max compared with maximum SUV of tumor. The vertical axis indicates the percentage volume difference between the two SUV thresholding methods.
724
I. J. Radiation Oncology
●
Biology
●
Physics
Fig. 6. Relationship of maximum tumor standard uptake value (SUV) vs. volume of tumor for gross tumor volumes (GTVs) contoured using SUV ⬎ 2.5 (GTV25) and SUV 40% Maximum SUV (GTV40). The lines represent a linear regression between the maximum tumor SUV and volume for the respective thresholding methods.
The mean percent change from the GTV CT was ⫺260% (median, ⫺20%; range, ⫺1,960% to ⫹40%) for GTV 2.5 and a mean percentage change of ⫺162% (median, ⫺140%; range, ⫺630% to 70%) for GTV 40% Max. When a boxwhisker analysis is performed to visualize the distribution of tumor volumes according to technique (Fig. 7a), as well as a separate plot of the PET technique volume difference from the GTV CT (Fig. 7b), there is stronger correlation of volumes between the GTV CT and GTV25. The values that are outliers in this analysis were those patients with large areas of atelectasis on CT with significant reduction in GTV based on either PET thresholding method.
Volume 67, Number 3, 2007
could have been overtreated, both scenarios in which FDGPET has a documented advantage over CT (29, 30). Theoretically, more accurate delineation of the GTV would result in more conformal target coverage, improved sparing of normal tissue, and ultimately enhanced local control. The optimal PET volume for radiation therapy, however, has yet to be defined. As stated, radiation oncologists have used a variety of methods from qualitative visualization, 40% intensity level, 42% intensity level, 50% intensity level relative to tumor SUV maximum, to a method involving a systematic linear regression function, or simply contouring any value of SUV greater than 2.5 (12, 16 –21). Regardless of what method of visualizing SUV is used, it is clear that the incorporation of FDG-PET results in alterations in radiotherapy treatment volumes. In our data, the significance of the change in volume from the CT-based target varied greatly with the SUV threshold used. Using a SUV 2.5 threshold resulted in an alteration of volume in 58% of our patients, which is in concordance with previously published data (12). However, using a SUV threshold of 40% of maximum tumor SUV resulted in a significant alteration in 95% of our patients, all of whom had a decrease in the tumor contoured. While the use of FDG-PET in addition to CT can improve concordance between radiation oncologists, especially in cases of atelectasis (29), our data show considerable differences in volume depending on the thresholding method used. However, our clinical data do suggest there may be a volume and maximal SUV threshold that negates target volume differences between SUV thresholding techniques. A novel aspect of this analysis is the inclusion of method
DISCUSSION In the comprehensive management of patients with NSCLC, FDG-PET plays a fundamental role in staging, evaluation of response to therapy, and more recently radiation therapy treatment planning (12, 24, 25). Improvement in patient selection for definitive radiotherapy will ultimately translate into improved outcomes as a result of more accurate staging. Published series have reported 8 –18% detection of distant metastases by PET at the time of radical radiotherapy planning, changing the intent of therapy from curative to palliative (12, 26). Moreover, the use of PET to assess “metabolic response” after therapy has been shown to be a surrogate marker for local control and prognosis (26 – 28). While it is clear that FDG-PET impacts patient care, much of our current understanding of the use of definitive radiotherapy in NSCLC is derived from clinical trials based on CT imaging without PET scans. It is conceivable that a significant number of patients in these studies could have had distant metastases. Furthermore, with CT-based target delineation, it is possible that patients with undetected involved small mediastinal lymph nodes could have been undertreated, and conversely, patients with tumors in the background of atelectasis without distinct tumor boundaries
Fig. 7. (a) Box-and-whisker distribution of gross tumor volumes (GTVs) contoured according to computed tomography (GTV CT), positron emission tomography (PET) using standard uptake values (SUVs) ⬎2.5 (GTV2.5), and PET using SUVs 40% Maximum SUV (GTV40%). (b) Box-and-whisker distribution of the percentage change from GTV volume based on CT. 2.5 SUV represents the percent difference between GTVs based on CT vs. PET SUV ⬎2.5. Similarly, SUV 40% Max represents the percent difference between GTVs based on CT vs. PET SUV of 40%Maximum SUV.
Correlation of PET standard uptake value and CT window-level thresholds
Fig. 8. Representation of regions of standard uptake value (SUV) using our visualization scheme. This patient had a maximum SUV of 9.3. The first magenta volume represents SUV regions of ⬎9, and each subsequent picture is ⬎8, ⬎7, ⬎6, ⬎5, ⬎4, ⬎3, ⬎2.5. The final two pictures represent the contours of each corresponding SUV threshold. Theoretically, a “metabolic boost” volume can be treated differentially based on SUV using our visualization technique.
of correlating CT-based window and level values to FDGPET SUV. Regardless of the thresholding method used, our correlation allows enhanced visualization of SUV regions of interest, ultimately resulting in less interobserver and intraobserver variability. We have developed a website (http://www.cuberick.com/⬃faisalv/window-level-suv-correlator. html) where users can convert PET SUV thresholds conveniently to CT window and level values. This site also provides values that “sharpen” the SUV edges by narrowing the window range. With the ability to segment regions of varying SUVs, the possibility exists to differentially target areas based on metabolic activity (Fig. 8). As Paulino and Johnstone point out in their editorial (31), a consensus opinion among radiation oncologists still has not been achieved. Although we recommend the use of an SUV cutoff of 2.5 based on our data, it is our institutional practice to confer with a nuclear medicine physician, and
●
R. HONG et al.
725
consider the clinical picture of the patient in toto when developing a radiation therapy treatment plan. In addition, we are well aware of the multitude of factors that affect tracer uptake, including the amount of dose administered, route of tracer administration, blood perfusion of tissue of interest, physiologic condition of the body (e.g., insulin resistance, body fat content), time of uptake determination after tracer administration, among others (16), and the errors introduced in image fusion and respiratory motion. The motion of lung cancer with breathing is complex and subject to intra- and interpatient variation. With the emergence of combined PET/CT scanners, the CT scan performed at the time of PET could be registered to the planning CT scan, facilitating the planning process. The hardware fusion of the PET/CT may mitigate errors attributable to image registration during RT planning. While it is not the focus of this study, the aforementioned variables have been standardized at our institution and are awaiting further analysis.
CONCLUSIONS Visualization of FDG-PET images in CT-based RT treatment planning requires more precision because the viewing W/Ls are less well established and defined compared with CT. Using the method of correlating SUV to CT Windows and Levels thresholds permits accurate displaying of PET data in registered PET/CT studies, allowing enhanced target delineation. However the optimal way to incorporate the PET SUV thresholds to contour GTV depends on the maximum tumor SUV and volume. At SUV values ⬎7, there is concordance in the size of the tumor volume between the thresholding methods employed in our study. For regions of SUV 4 – 6 the best correlation occurs for tumors of small volume near 100 cc. Due to tumor heterogeneity, and the wide variability in volumes obtained by using SUV 40% Max, we recommend using areas of SUV ⱖ 2.5 with the aid of a physician with specialized training in interpreting PET scans for RT planning purposes in NSCLC.
REFERENCES 1. Dillman RO, Herndon J, Seagren SL, et al. Improved survival in stage III NSCLC: Seven year follow-up of CALGB 8433 trial. J Natl Cancer Inst 1996;88:1210 –1215. 2. Furuse K, Fukuoka M, Kawahara M, et al. Phase III study of concurrent versus sequential thoracic radiotherapy in combination with mitomycin, vindesine, and cisplatin in unresectable stage III non–small cell lung cancer. J Clin Oncol 1999; 17:2692–2699. 3. Lau DH, Crowley JJ, et al. Southwest Oncology Group phase II trial of concurrent carboplatin, etoposide and radiation for poor-risk stage III non–small cell lung cancer. J Clin Oncol 1998;16:3078 –3081. 4. Sawyer TE, Bonner JA. Unresectable or medically inoperable NSCLC: The use of established clinical prognostic factors in making radiation-related treatment decisions. Semin Radiat Oncol 2000;10:267–273. 5. Armstrong J, Raben A, Zelefsky M, et al. Promising survival
6.
7.
8. 9.
10.
with three-dimensional conformal radiation therapy for non–small cell lung cancer. Radiother Oncol 1997;44:17–22. Rengan R, Rosenzweig KE, Venkatraman E, et al. Improved local control with higher doses of radiation in large-volume Stage III NSCLC. Int J Radiat Oncol Biol Phys 2004;60: 741–747. Narayan S, Henning GT, Ten Haken RK, et al. Results following treatment to doses of 92.4 or 102.9 Gy on a phase I dose escalation study for non–small cell lung cancer. Lung Cancer 2004;44:79 – 88. Sanjiv S, Gambhir, Czernin J, et al. A tabulated summary of the FDG PET literature. J Nucl Med 2001;42:1–93. Sasaki M, Ichiya Y, Kuwabara Y, et al. The usefulness of FDG PET for the detection of mediastinal lymph node metastases in patients with NSCLC: A comparative study with X-ray computed tomography. Eur J Nucl Med 1996;23:741–747. Bayne M, Everitt S, MacManus M, et al. Fused PET/CT
726
11.
12. 13.
14. 15.
16. 17.
18. 19. 20. 21.
22.
I. J. Radiation Oncology
●
Biology
●
Physics
images for planning radical radiotherapy for NSCLC. Lung Cancer 2005;49(Suppl. 3):S74 –S75. MacManus MP, Hicks R, Matthews J, et al. Long-term follow-up of 153 candidates for definitive radiation or chemoradiation therapy with non–small cell lung cancer who were staged by PET: Implications for design of future randomized trials. Int J Radiat Oncol Biol Phys 2005;63(Suppl. 1):S223–S224. Bradley J, Thorstad W, Mutic S, et al. Impact of FDG-PET on radiation therapy volume delineation in NSCLC. Int J Radiat Oncol Biol Phys 2004;59:78 – 86. Mah K, Caldwell C, Ung Y, et al. The impact of FDG-PET on target and critical organs in CT based treatment planning of patients with poorly defined NSCLC: A prospective study. Int J Radiat Oncol Biol Phys 2002;52:339 –350. Giraud P, Grahek D, Montravers F, et al. CT and FDG image fusion for optimization of conformal radiotherapy of lung cancers. Int J Radiat Oncol Biol Phys 2001;49:1249 –1257. Steenbakkers R, Duppen J, Fitton I, et al. Reduction of observer variation using matched CT-PET for lung cancer delineation: A three-dimensional analysis. Int J Radiat Oncol Biol Phys 2006;64:435– 448. Huang SC. Anatomy of SUV. Standardized uptake value. Nucl Med Biol 2000;27:643– 646. Vansteenkiste JF, Stroobants SG, Dupont PJ, et al. Prognostic importance of the standardized uptake value on FDG PET in NSCLC: An analysis of 125 cases. J Clin Oncol 1999;17: 3201–3206. Erdi Y, Rosenzweig K, Erdi A, et al. Radiotherapy treatment planning for patients with NSCLC using PET. Radiother Oncol 2002;62:51– 60. Erdi YE, Mawlawi O, Larson SM, et al. Segmentation of lung lesion volume by adaptive positron emission tomography image thresholding. Cancer 1997;80(12 Suppl.):2505–2509. Black QC, Grills IS, Kestin LL, et al. Defining a radiotherapy target with positron emission tomography. Int J Radiat Oncol Biol Phys 2004;60:1272–1282. Caldwell CB, Mah K, Ung YC, et al. Observer variation in contouring gross tumour volume in patients with poorly defined NSCLC on CT: The impact of FDG PET fusion. Int J Radiat Oncol Biol Phys 2001;51:923–931. Van de Steene J, Linthout N, de Mey J, et al. Definition of
Volume 67, Number 3, 2007
23.
24.
25.
26.
27.
28.
29.
30.
31.
gross tumor volume in lung cancer: Inter-observer variability. Radiother Oncol 2002;62:37– 49. Giraud P, Elles S, Helfre S, et al. Conformal radiotherapy for lung cancer: Different delineation of the gross tumor volume (GTV) by radiologists and radiation oncologists. Radiother Oncol 2002;62:27–36. MacManus MP, Hicks RJ, Matthews JP, et al. Positron emission tomography is superior to computed tomography scanning for response assessment after radical radiotherapy or chemoradiotherapy in patients with non–small cell lung cancer. J Clin Oncol 2003;21:1285–1292. Lardinois D, Weder W, Hany TF, et al. Staging of non–smallcell lung cancer with integrated positron-emission tomography and computed tomography. N Engl J Med 2003;348: 2500 –2507. MacManus MP, Hicks RJ, Ball DL, et al. F-18 fluorodeoxyglucose positron emission tomography staging in radical radiotherapy candidates with non small cell lung carcinoma: Powerful correlation with survival and high impact on treatment. Cancer 2001;92:886 – 895. Vansteenkiste JF, Stroobants SG, DeLyn PR, et al. Potential use of FDG-PET scan after induction chemotherapy in surgically staged NSCLC: A prospective pilot study. The Leuven Lung Cancer Group. Ann Oncol 1998;9:1193–1198. Vansteenkiste JF, Stroobants S, Hoekstra C, et al. 18fluoro2-deoxyglucose positron emission tomography (PET) in the assessment of induction chemotherapy (IC) in stage IIIA-N2 NSCLC: A multi-center prospective study. Proc Am Soc Clin Oncol 2001;20:313A. Nestle U, Walter K, Schmidt S, et al. FDG-PET for the planning of radiotherapy in lung cancer: High impact in patients with atelectasis. Int J Radiat Oncol Biol Phys 1999;44: 593–597. Vanuytsel LJ, Vansteenkiste JF, Stroobants SG, et al. The impact of FDG-PET lymph node staging on the radiation treatment volumes in patients with non–small cell lung cancer. Radiother Oncol 2000;55:317–324. Paulino AC, Johnstone P. FDG-PET in radiotherapy treatment planning: Pandora’s box? Int J Radiat Oncol Biol Phys 2004; 59:4 –5.
doi:10.1016/j.ijrobp.2006.09.039
CLINICAL INVESTIGATION
Lung
CORRELATION OF PET STANDARD UPTAKE VALUE AND CT WINDOW-LEVEL THRESHOLDS FOR TARGET DELINEATION IN CT-BASED RADIATION TREATMENT PLANNING ROBERT HONG, M.D.,* JAMES HALAMA, PH.D.,† DAVIDE BOVA, M.D.,† ANIL SETHI, PH.D.,* AND BAHMAN EMAMI, M.D.* *Department of Radiation Oncology, and †Department of Radiology, Nuclear Medicine Division, Loyola University Medical Center, Maywood, IL Purpose: To develop standardized correlates of [18F]fluoro-2-deoxy-D-glucose positron emission tomography (FDG-PET) standard uptake value (SUV) to computed tomography (CT)-based window and levels. Methods and Materials: Nineteen patients with non–small-cell lung cancer who underwent imaging with positron emission tomography (PET) and CT were selected. A method of standardizing SUV within CT planning software was developed. A scale factor, determined by a sensitivity calibration of the PET scanner, converts voxel counts to activity per gram in tissue, allowing SUVs to be correlated to CT window and levels. A method of limiting interobserver variations was devised to enhance “edges” of regions of interest based on SUV thresholds. The difference in gross tumor volumes (GTVs) based on CT, PET SUV > 2.5, and regions of 40% maximum SUV were analyzed. Results: The mean SUV was 9.3. Mean GTV volumes were 253 cc for CT, 221 cc for SUV > 2.5, and 97 cc for SUV40%Max. Average volume difference was ⴚ259% between >2.5 SUV and CT and ⴚ162% between SUV40%Max and CT. Percent difference between GTV > 2.5 SUV and SUV40%Max remained constant beyond SUV > 7. For SUVs 4 – 6, best correlation among SUV thresholds occurred at volumes near 90 cc. Mean percent change from GTVs contoured according to CT (GTV CT) was ⴚ260% for GTV2.5 and ⴚ162% for GTV40%Max. Using the SUV40%Max threshold resulted in a significant alteration of volume in 98% of patients, while the SUV2.5 threshold resulted in an alteration of volume in 58% of patients. Conclusions: Our method of correlating SUV to W/L thresholds permits accurate displaying of SUV in coregistered PET/CT studies. The optimal SUV thresholds to contour GTV depend on maximum tumor SUV and volume. Best correlation occurs with SUVs >6 and small volumes <100 cc. At SUVs >7, differences between the SUV threshold filters remain constant. Because of variability in volumes obtained by using SUV40%Max, we recommend using SUV > 2.5 for radiotherapy planning in non–small-cell lung cancer. © 2007 Elsevier Inc. Radiotherapy, Positron emission tomography (PET), Standard uptake value (SUV), Target delineation, NSCLC.
INTRODUCTION
modulated RT, stereotactic radiosurgery, and image-guided RT depend on precise tumor volume delineation, which is again dependent on radiologic imaging. Although computed tomography (CT) imaging techniques provide anatomic information in addition to electron densities of various tissues necessary for dose distribution information in computerized treatment planning systems, the advantage over other modalities is relative, especially in distinguishing malignancy from normal tissues in the context of atelectasis or pleural effusion. [18F]fluoro-2-deoxy-D-glucose positron emission tomography (FDG-PET) is a functional imaging modality that is increasingly being used in all aspects of oncology, including
Patients with unresectable or medically inoperable non– small-cell lung cancer (NSCLC) have a poor prognosis. These patients who receive radiotherapy (RT) with or without chemotherapy have a 5-year survival between 0% and 30% (1– 4). Although largely dependent on tumor stage and performance status, published literature suggests tumor control is improved with dose escalation using conformal radiation therapy (5–7). New treatment techniques are increasingly being used to allow additional dose escalation beyond those achieved by conventional three-dimensional conformal radiotherapy. Treatment modalities such as intensity-
Society for Therapeutic Radiology and Oncology (ASTRO), in Philadelphia, PA, November 5–9, 2006. Conflict of interest: none. Received Aug 14, 2006, and in revised form Sept 20, 2006. Accepted for publication Sept 21, 2006.
Reprint requests to: Bahman Emami, M.D., Loyola University Medical Center, Department of Radiation Oncology, 2160 S. 1st Avenue, Maguire Building, Room 2944, Maywood, IL 60153. Tel: (708) 216-2555; Fax: (708) 216-6076; E-mail: [email protected] Presented in part at the 48th Annual Meeting of the American 720
Correlation of PET standard uptake value and CT window-level thresholds
diagnostic evaluation of primary lung nodules, staging of mediastinal lymph nodes, and detection of distant metastases. In an overview of the available literature, FDG-PET was found to have a 91% sensitivity, 68% specificity in diagnosing primary lung cancer, a 83% sensitivity and 91% specificity in mediastinal staging (8), compared with 56 – 65% sensitivity and 73– 87% specificity for mediastinal staging for CT (9). MacManus et al. reported in an interim analysis of a prospective protocol evaluating radical radiotherapy for NSCLC, that nearly a third (5/18) of patients were excluded owing to the identification of metastatic disease or extensive intrathoracic disease by PET. Additionally, in half (6/12) of those proceeding to RT planning, ⬎25% of the PET/CT gross tumor volume (GTV) was not covered by the CT GTV (10, 11). Fused images of PET and CT studies have the cumulative benefit of providing physiologic data with precise topographic localization. Although there are reports of initial studies incorporating FDG-PET into RT treatment planning, there is no consensus on the use of standard uptake value (SUV) in target delineation (12–15). For lung nodules, in addition to size, shape, and location, an SUV ⬎2.5 is a commonly used threshold indicative of malignancy (16, 17). Others have used a preclinical procedure described by Erdi et al. that used phantom studies to accurately define lesions by setting the upper window level to the maximum intensity voxel and the lower level set at a value of 42% of the upper level for spherical lesions of ⬎4 cc (18, 19). Black et al. also used phantom studies to describe a method of defining GTV based on a linear regressive function threshold of SUV that resulted in smaller deviation when compared with a fixed image intensity threshold (42% of maximum intensity) (20). It is clear that with PET-defined tumor volumes in RT treatment planning, variation in setting image signal thresholds can significantly affect the contour of the GTV, resulting in considerable interobserver and intraobserver variation (21–23). Moreover, there is no reliable correlation of CT window and level settings to PET SUV such that PET metabolic data in the form of SUV can be consistently viewed in CT-based treatment planning software. A retrospective analysis investigating the impact of GTV delineation for NSCLC according to varying PET SUV thresholds is described. Fused image sets from separately acquired FDG-PET with CT simulation were used. This study provides a novel method of correlating CT window and level values to PET SUV values, for precise visualization of PET data. Furthermore, by modifying our visualization technique, we have developed a means to filter PET data to segment regions of interest based on user-defined SUV boundaries, effectively sharpening the tumor edges. The goal of this study is to describe a practical method for GTV delineation based on FDG-PET SUV visualization to aid radiation oncologists in image-based RT treatment planning.
●
R. HONG et al.
721
METHODS AND MATERIALS Study population This study was approved by Loyola University Medical Center’s Institutional Review Board. Each patient was required to have pathologic confirmation of NSCLC, locally advanced unresectable or medically inoperable with a Karnofsky performance status of 70 –100. Nineteen consecutive patients who underwent staging PET evaluation between December 2005 and June 2006, who also were planned for CT-based three-dimensional conformal radiotherapy, were included in the current analysis.
Patient simulation Patients were CT simulated supine with arms overhead with upper body custom alpha cradle immobilization (Smithers Medical Products, North Canton, OH). The alpha cradle was fabricated to fit within the PET scanner aperture, which required a smaller width and flat table top. Five CT and PET compatible fiducial markers were also placed in noncoplanar positions on the cradle to aid in coregistration of image sets in the axial, sagittal, and coronal planes. Fiducial markers consisted of radiopaque “donuts” which housed FDG-soaked gauze at the time of PET scan. An isocenter was localized for each patient. After a scout study, the patient was given 1 tablespoon of barium paste by mouth and 100 cc of intravenous contrast. The scan was performed immediately using a Philips AcQSim CT scanner (Philips Medical Systems N.A., Bothell, WA) acquiring images at 5 mm ⫻ 5 mm intervals (512 ⫻ 512 matrix, 0.94 mm pixel size) from the second cervical vertebral body to the liver over 45 s. The patients were scanned while breathing freely (Fig. 1). Patients then proceeded for PET scan with the compatible immobilization devices within 24 – 48 h of CT scan acquisition. The patients were scanned with a dedicated whole-body Philips
Fig. 1. (a) Patients were simulated with computed tomography (CT) and positron emission tomography (PET) within 72 h. Custom fabricated immobilization was the same for each scan. (b) Four noncoplanar fiducials were placed on the alpha cradle. At the time of positron emission tomography simulation, [18F]fluoro-2deoxy-D-glucose was embedded in the center, allowing for improved image registration.
722
I. J. Radiation Oncology
●
Biology
●
Physics
Volume 67, Number 3, 2007
Fig. 2. (a) Schematic representation of positron emission tomography standard uptake value and computed tomography window and level (voxel) relationship. (b) Narrowing the displayed window to filter [18F]fluoro-2-deoxy-D-glucose positron emission tomography image to allow enhanced visualization of volume edges.
Allegro System PET camera. All patients had their height and weight determined at the time of study. Four-hour resting glucose values were measured and determined to be within acceptable ranges (⬎65 to ⬍200). Intravenous access was obtained. FDG quantity was based on weight and ranged between 6 mCi and 14.5 mCi. With gentle intravenous fluid hydration, FDG was injected intravenously and isotope was distributed over 90 min in a resting state. The patient was positioned and fiducials were prepared with gauze-soaked FDG. Scan acquisition occurred over 45 min.
PET interpretation The PET study was qualitatively reviewed by an experienced nuclear radiologist in comparison with prior diagnostic chest CT studies. Generally, regions of increased and abnormal FDG uptake beyond expected normal tissue uptake were characterized as tumor when SUV ⬎ 2.5. However, this qualitative assessment took all clinical and imaging factors into consideration including size and location of the lesion, histologic differentiation of the tumor, and inflammatory processes. For example, a region of “streaking” SUV ⬎ 2.5 was detected in the supraclavicular fossa of a patient that was qualitatively interpreted as a benign inflammatory response. Likewise, a malignant-appearing 1-cm nodule evaluable on CT but with SUV ⬍ 2.5 was scored as tumor. This qualitative binary assessment of the PET study was performed in the presence of the treating radiation oncologist and considered in the final gross tumor target delineation. In PET, the raw data corresponds to a count of the number of positron annihilations in a voxel. A scan-specific scale factor, determined by a sensitivity calibration (SUVcal) of the PET raw data, is used to convert voxel counts to activity per gram in the tissue. The ratio between the SUVcal (i.e., raw data) and SUV for any voxel in the PET image is constant. The SUV is then the ratio of the activity per gram in the tumor to the administered activity per gram of total body mass (Fig. 2). Both the maximum tumor SUV and SUVcal for the PET study were recorded.
PET data were initially visualized using arbitrary window and level values defined subjectively to ensure proper image registration. Then, based on the maximum tumor SUV and the minimum threshold SUV, window and level values were generated based on the PET sensitivity calibration factor. The minimum SUV thresholds that were evaluated were regions ⬎2.5 and 40% maximum SUVs based on published literature recommendations (16 –19). Although this initial filter allowed qualitative visualization of SUV of the full range of activity defined by the minimum and maximum thresholds, the “edges” of the uptake were not clearly defined. By narrowing the displayed window values to within 0.1 of the minimum SUV, the regions at the periphery of the area of interest could be sharpened (Fig. 3). This allowed for precise contouring of the user-defined regions based on the available PET data (Fig. 4). The CT alone tumor volumes consisted of the primary tumor and any regional lymph nodes with a diameter of ⬎1 cm on the short axis. The window width and level in the CT images were 1,600 Hounsfield units (HU) and ⫺600 HU for the lung parenchyma and 400 HU and 20 HU for the mediastinum. Total tumor volume for the GTVs contoured using CT alone (GTV CT), PET SUV ⬎ 2.5 (GTV25), and PET SUV 40% maximum (GTV40) were generated using XIO planning software.
Statistical analysis The GTV volumes generated by the two SUV thresholds (GTV25, GTV40) were compared with each other as well as the
Target delineation Positron emission tomography Digital Imaging and Communications in Medicine (DICOM) data were sent to the radiotherapy planning software (CMS, Inc. Focal and XIO v4.3.1) where a combination of automated fusion using Focal’s Mutual Information algorithm and manual registration of overlapping fiducials was performed. Maximum tumor SUV values and the scale factor, determined by a scan-specific sensitivity calibration of the PET image, were used to convert voxel counts to SUV (Fig. 2). The
Fig. 3. Illustration of the variables and formulas used to generate window and level values to display standard uptake value (SUV) in the computed tomography– based radiotherapy planning software. (Available online at http://www.cuberick.com/⬃faisalv/ window-level-suv-correlator.html).
Correlation of PET standard uptake value and CT window-level thresholds
●
R. HONG et al.
723
Fig. 4. Example of gross tumor that is difficult to visualize on computed tomography with stepwise visualization of gross tumor volume based on coregistered positron emission tomography scan: (a) represents fused images displayed with a full range of standard uptake value (SUV); (b) represents fused images only displaying regions between SUV 2.5 and SUV max; (c) represents fused images with a narrowed window displaying SUV 2.5–2.6, resulting in sharpened edges for the regions of interest.
GTV CT. The percentage difference between GTV25 and GTV40 was calculated. The percentage difference between GTV25 and GTV CT as well as GTV40 and GTV CT was also calculated based on absolute volumes. Linear regression analysis, box-whisker plot visualization and graphs were generated by Systat (Version 11; Systat Software Inc., Point Richmond, CA).
RESULTS The mean SUV for the tumors examined was 9.3 (median, 9.0, range, 4.4 –14.8). The mean GTV volumes were 253 cc (range, 12–922 cc) on CT, 221 cc (range, 1–920 cc) for SUV ⱖ 2.5 and 97 cc (range, 4 –324 cc) for SUV 40%Max. The average volume difference was ⫺259% between the ⱖ2.5 SUV and CT GTV and ⫺162% between the SUV 40% Max and CT GTV. Overall, using a 25% volume difference as a cutoff for significant alteration of GTV CT, 58% (11/19) of the GTV25 and 95% (18/19) of GTV40 volumes were substantially different from the original CTbased tumor volumes. The two largest differences occurred in patients whose tumor was delineated within atelectatic lung parenchyma. An analysis between the two PET thresholding techniques was also performed. When the percentage difference between the tumor volumes is plotted against their respective maximum SUV (Fig. 5), the percent difference between GTV ⱖ 2.5 SUV and SUV 40% Max remains constant after SUV ⬎ 7, with the GTV 40% Max volume averaging ⫹ 55%. There is a clear plateau on the shape of this curve which denotes minimal difference in volume between the
two SUV thresholding methods at very high SUV values. Conversely, there is a steep slope denoting considerable variation in tumor volumes between the two techniques at nominal SUVs between 4 and 6. To further analyze whether there was any correlation with tumor volume, the GTV volumes as determined by the two thresholds were plotted against the maximum SUV for each respective patient. Using a linear regression analysis, the best correlation of GTV among the SUV threshold values occurs at small tumor volumes near 90 –100 cc for SUVs between 4 – 6 (Fig. 6).
Fig. 5. Relationship of the percentage difference between the gross tumor volume (GTV) contoured based on a standard uptake value (SUV) ⬎2.5 vs. SUV 40% Max compared with maximum SUV of tumor. The vertical axis indicates the percentage volume difference between the two SUV thresholding methods.
724
I. J. Radiation Oncology
●
Biology
●
Physics
Fig. 6. Relationship of maximum tumor standard uptake value (SUV) vs. volume of tumor for gross tumor volumes (GTVs) contoured using SUV ⬎ 2.5 (GTV25) and SUV 40% Maximum SUV (GTV40). The lines represent a linear regression between the maximum tumor SUV and volume for the respective thresholding methods.
The mean percent change from the GTV CT was ⫺260% (median, ⫺20%; range, ⫺1,960% to ⫹40%) for GTV 2.5 and a mean percentage change of ⫺162% (median, ⫺140%; range, ⫺630% to 70%) for GTV 40% Max. When a boxwhisker analysis is performed to visualize the distribution of tumor volumes according to technique (Fig. 7a), as well as a separate plot of the PET technique volume difference from the GTV CT (Fig. 7b), there is stronger correlation of volumes between the GTV CT and GTV25. The values that are outliers in this analysis were those patients with large areas of atelectasis on CT with significant reduction in GTV based on either PET thresholding method.
Volume 67, Number 3, 2007
could have been overtreated, both scenarios in which FDGPET has a documented advantage over CT (29, 30). Theoretically, more accurate delineation of the GTV would result in more conformal target coverage, improved sparing of normal tissue, and ultimately enhanced local control. The optimal PET volume for radiation therapy, however, has yet to be defined. As stated, radiation oncologists have used a variety of methods from qualitative visualization, 40% intensity level, 42% intensity level, 50% intensity level relative to tumor SUV maximum, to a method involving a systematic linear regression function, or simply contouring any value of SUV greater than 2.5 (12, 16 –21). Regardless of what method of visualizing SUV is used, it is clear that the incorporation of FDG-PET results in alterations in radiotherapy treatment volumes. In our data, the significance of the change in volume from the CT-based target varied greatly with the SUV threshold used. Using a SUV 2.5 threshold resulted in an alteration of volume in 58% of our patients, which is in concordance with previously published data (12). However, using a SUV threshold of 40% of maximum tumor SUV resulted in a significant alteration in 95% of our patients, all of whom had a decrease in the tumor contoured. While the use of FDG-PET in addition to CT can improve concordance between radiation oncologists, especially in cases of atelectasis (29), our data show considerable differences in volume depending on the thresholding method used. However, our clinical data do suggest there may be a volume and maximal SUV threshold that negates target volume differences between SUV thresholding techniques. A novel aspect of this analysis is the inclusion of method
DISCUSSION In the comprehensive management of patients with NSCLC, FDG-PET plays a fundamental role in staging, evaluation of response to therapy, and more recently radiation therapy treatment planning (12, 24, 25). Improvement in patient selection for definitive radiotherapy will ultimately translate into improved outcomes as a result of more accurate staging. Published series have reported 8 –18% detection of distant metastases by PET at the time of radical radiotherapy planning, changing the intent of therapy from curative to palliative (12, 26). Moreover, the use of PET to assess “metabolic response” after therapy has been shown to be a surrogate marker for local control and prognosis (26 – 28). While it is clear that FDG-PET impacts patient care, much of our current understanding of the use of definitive radiotherapy in NSCLC is derived from clinical trials based on CT imaging without PET scans. It is conceivable that a significant number of patients in these studies could have had distant metastases. Furthermore, with CT-based target delineation, it is possible that patients with undetected involved small mediastinal lymph nodes could have been undertreated, and conversely, patients with tumors in the background of atelectasis without distinct tumor boundaries
Fig. 7. (a) Box-and-whisker distribution of gross tumor volumes (GTVs) contoured according to computed tomography (GTV CT), positron emission tomography (PET) using standard uptake values (SUVs) ⬎2.5 (GTV2.5), and PET using SUVs 40% Maximum SUV (GTV40%). (b) Box-and-whisker distribution of the percentage change from GTV volume based on CT. 2.5 SUV represents the percent difference between GTVs based on CT vs. PET SUV ⬎2.5. Similarly, SUV 40% Max represents the percent difference between GTVs based on CT vs. PET SUV of 40%Maximum SUV.
Correlation of PET standard uptake value and CT window-level thresholds
Fig. 8. Representation of regions of standard uptake value (SUV) using our visualization scheme. This patient had a maximum SUV of 9.3. The first magenta volume represents SUV regions of ⬎9, and each subsequent picture is ⬎8, ⬎7, ⬎6, ⬎5, ⬎4, ⬎3, ⬎2.5. The final two pictures represent the contours of each corresponding SUV threshold. Theoretically, a “metabolic boost” volume can be treated differentially based on SUV using our visualization technique.
of correlating CT-based window and level values to FDGPET SUV. Regardless of the thresholding method used, our correlation allows enhanced visualization of SUV regions of interest, ultimately resulting in less interobserver and intraobserver variability. We have developed a website (http://www.cuberick.com/⬃faisalv/window-level-suv-correlator. html) where users can convert PET SUV thresholds conveniently to CT window and level values. This site also provides values that “sharpen” the SUV edges by narrowing the window range. With the ability to segment regions of varying SUVs, the possibility exists to differentially target areas based on metabolic activity (Fig. 8). As Paulino and Johnstone point out in their editorial (31), a consensus opinion among radiation oncologists still has not been achieved. Although we recommend the use of an SUV cutoff of 2.5 based on our data, it is our institutional practice to confer with a nuclear medicine physician, and
●
R. HONG et al.
725
consider the clinical picture of the patient in toto when developing a radiation therapy treatment plan. In addition, we are well aware of the multitude of factors that affect tracer uptake, including the amount of dose administered, route of tracer administration, blood perfusion of tissue of interest, physiologic condition of the body (e.g., insulin resistance, body fat content), time of uptake determination after tracer administration, among others (16), and the errors introduced in image fusion and respiratory motion. The motion of lung cancer with breathing is complex and subject to intra- and interpatient variation. With the emergence of combined PET/CT scanners, the CT scan performed at the time of PET could be registered to the planning CT scan, facilitating the planning process. The hardware fusion of the PET/CT may mitigate errors attributable to image registration during RT planning. While it is not the focus of this study, the aforementioned variables have been standardized at our institution and are awaiting further analysis.
CONCLUSIONS Visualization of FDG-PET images in CT-based RT treatment planning requires more precision because the viewing W/Ls are less well established and defined compared with CT. Using the method of correlating SUV to CT Windows and Levels thresholds permits accurate displaying of PET data in registered PET/CT studies, allowing enhanced target delineation. However the optimal way to incorporate the PET SUV thresholds to contour GTV depends on the maximum tumor SUV and volume. At SUV values ⬎7, there is concordance in the size of the tumor volume between the thresholding methods employed in our study. For regions of SUV 4 – 6 the best correlation occurs for tumors of small volume near 100 cc. Due to tumor heterogeneity, and the wide variability in volumes obtained by using SUV 40% Max, we recommend using areas of SUV ⱖ 2.5 with the aid of a physician with specialized training in interpreting PET scans for RT planning purposes in NSCLC.
REFERENCES 1. Dillman RO, Herndon J, Seagren SL, et al. Improved survival in stage III NSCLC: Seven year follow-up of CALGB 8433 trial. J Natl Cancer Inst 1996;88:1210 –1215. 2. Furuse K, Fukuoka M, Kawahara M, et al. Phase III study of concurrent versus sequential thoracic radiotherapy in combination with mitomycin, vindesine, and cisplatin in unresectable stage III non–small cell lung cancer. J Clin Oncol 1999; 17:2692–2699. 3. Lau DH, Crowley JJ, et al. Southwest Oncology Group phase II trial of concurrent carboplatin, etoposide and radiation for poor-risk stage III non–small cell lung cancer. J Clin Oncol 1998;16:3078 –3081. 4. Sawyer TE, Bonner JA. Unresectable or medically inoperable NSCLC: The use of established clinical prognostic factors in making radiation-related treatment decisions. Semin Radiat Oncol 2000;10:267–273. 5. Armstrong J, Raben A, Zelefsky M, et al. Promising survival
6.
7.
8. 9.
10.
with three-dimensional conformal radiation therapy for non–small cell lung cancer. Radiother Oncol 1997;44:17–22. Rengan R, Rosenzweig KE, Venkatraman E, et al. Improved local control with higher doses of radiation in large-volume Stage III NSCLC. Int J Radiat Oncol Biol Phys 2004;60: 741–747. Narayan S, Henning GT, Ten Haken RK, et al. Results following treatment to doses of 92.4 or 102.9 Gy on a phase I dose escalation study for non–small cell lung cancer. Lung Cancer 2004;44:79 – 88. Sanjiv S, Gambhir, Czernin J, et al. A tabulated summary of the FDG PET literature. J Nucl Med 2001;42:1–93. Sasaki M, Ichiya Y, Kuwabara Y, et al. The usefulness of FDG PET for the detection of mediastinal lymph node metastases in patients with NSCLC: A comparative study with X-ray computed tomography. Eur J Nucl Med 1996;23:741–747. Bayne M, Everitt S, MacManus M, et al. Fused PET/CT
726
11.
12. 13.
14. 15.
16. 17.
18. 19. 20. 21.
22.
I. J. Radiation Oncology
●
Biology
●
Physics
images for planning radical radiotherapy for NSCLC. Lung Cancer 2005;49(Suppl. 3):S74 –S75. MacManus MP, Hicks R, Matthews J, et al. Long-term follow-up of 153 candidates for definitive radiation or chemoradiation therapy with non–small cell lung cancer who were staged by PET: Implications for design of future randomized trials. Int J Radiat Oncol Biol Phys 2005;63(Suppl. 1):S223–S224. Bradley J, Thorstad W, Mutic S, et al. Impact of FDG-PET on radiation therapy volume delineation in NSCLC. Int J Radiat Oncol Biol Phys 2004;59:78 – 86. Mah K, Caldwell C, Ung Y, et al. The impact of FDG-PET on target and critical organs in CT based treatment planning of patients with poorly defined NSCLC: A prospective study. Int J Radiat Oncol Biol Phys 2002;52:339 –350. Giraud P, Grahek D, Montravers F, et al. CT and FDG image fusion for optimization of conformal radiotherapy of lung cancers. Int J Radiat Oncol Biol Phys 2001;49:1249 –1257. Steenbakkers R, Duppen J, Fitton I, et al. Reduction of observer variation using matched CT-PET for lung cancer delineation: A three-dimensional analysis. Int J Radiat Oncol Biol Phys 2006;64:435– 448. Huang SC. Anatomy of SUV. Standardized uptake value. Nucl Med Biol 2000;27:643– 646. Vansteenkiste JF, Stroobants SG, Dupont PJ, et al. Prognostic importance of the standardized uptake value on FDG PET in NSCLC: An analysis of 125 cases. J Clin Oncol 1999;17: 3201–3206. Erdi Y, Rosenzweig K, Erdi A, et al. Radiotherapy treatment planning for patients with NSCLC using PET. Radiother Oncol 2002;62:51– 60. Erdi YE, Mawlawi O, Larson SM, et al. Segmentation of lung lesion volume by adaptive positron emission tomography image thresholding. Cancer 1997;80(12 Suppl.):2505–2509. Black QC, Grills IS, Kestin LL, et al. Defining a radiotherapy target with positron emission tomography. Int J Radiat Oncol Biol Phys 2004;60:1272–1282. Caldwell CB, Mah K, Ung YC, et al. Observer variation in contouring gross tumour volume in patients with poorly defined NSCLC on CT: The impact of FDG PET fusion. Int J Radiat Oncol Biol Phys 2001;51:923–931. Van de Steene J, Linthout N, de Mey J, et al. Definition of
Volume 67, Number 3, 2007
23.
24.
25.
26.
27.
28.
29.
30.
31.
gross tumor volume in lung cancer: Inter-observer variability. Radiother Oncol 2002;62:37– 49. Giraud P, Elles S, Helfre S, et al. Conformal radiotherapy for lung cancer: Different delineation of the gross tumor volume (GTV) by radiologists and radiation oncologists. Radiother Oncol 2002;62:27–36. MacManus MP, Hicks RJ, Matthews JP, et al. Positron emission tomography is superior to computed tomography scanning for response assessment after radical radiotherapy or chemoradiotherapy in patients with non–small cell lung cancer. J Clin Oncol 2003;21:1285–1292. Lardinois D, Weder W, Hany TF, et al. Staging of non–smallcell lung cancer with integrated positron-emission tomography and computed tomography. N Engl J Med 2003;348: 2500 –2507. MacManus MP, Hicks RJ, Ball DL, et al. F-18 fluorodeoxyglucose positron emission tomography staging in radical radiotherapy candidates with non small cell lung carcinoma: Powerful correlation with survival and high impact on treatment. Cancer 2001;92:886 – 895. Vansteenkiste JF, Stroobants SG, DeLyn PR, et al. Potential use of FDG-PET scan after induction chemotherapy in surgically staged NSCLC: A prospective pilot study. The Leuven Lung Cancer Group. Ann Oncol 1998;9:1193–1198. Vansteenkiste JF, Stroobants S, Hoekstra C, et al. 18fluoro2-deoxyglucose positron emission tomography (PET) in the assessment of induction chemotherapy (IC) in stage IIIA-N2 NSCLC: A multi-center prospective study. Proc Am Soc Clin Oncol 2001;20:313A. Nestle U, Walter K, Schmidt S, et al. FDG-PET for the planning of radiotherapy in lung cancer: High impact in patients with atelectasis. Int J Radiat Oncol Biol Phys 1999;44: 593–597. Vanuytsel LJ, Vansteenkiste JF, Stroobants SG, et al. The impact of FDG-PET lymph node staging on the radiation treatment volumes in patients with non–small cell lung cancer. Radiother Oncol 2000;55:317–324. Paulino AC, Johnstone P. FDG-PET in radiotherapy treatment planning: Pandora’s box? Int J Radiat Oncol Biol Phys 2004; 59:4 –5.