- Email: [email protected]
Target Volume Delineation in Oropharyngeal Cancer: Impact of PET, MRI, and Physical Examination
International Journal of
Radiation Oncology biology
physics
www.redjournal.org
Clinical Investigation: Head and Neck Cancer
Target Volume Delineation in Oropharyngeal Cancer: Impact of PET, MRI, and Physical Examination Anuradha Thiagarajan, M.D.,* Nicola Caria, M.D.,* Heiko Scho¨der, M.D.,y N. Gopalakrishna Iyer, M.D.,z Suzanne Wolden, M.D.,* Richard J. Wong, M.D.,z Eric Sherman, M.D.,x Matthew G. Fury, M.D.,x and Nancy Lee, M.D.* From the Departments of *Radiation Oncology, yRadiology, zHead and Neck Surgery, and xMedical Oncology, Memorial Sloan-Kettering Cancer Center, New York, NY Received Jan 12, 2011, and in revised form May 29, 2011. Accepted for publication May 31, 2011
Summary In the era of high precision radiotherapy, accurate target delineation is crucial. Sole utilization of computed tomography scans in gross tumor volume delineation for head and neck cancers is subject to significant interobserver variation. This paper demonstrates that magnetic resonance imaging and positron emission tomography add valuable complementary information, and that their combined use is recommended. In addition, it shows that thorough physical examination is invaluable in assessing superficial tumor extent in oropharyngeal malignancies, a dimension that is often missed or underestimated by imaging alone.
Introduction: Sole utilization of computed tomography (CT) scans in gross tumor volume (GTV) delineation for head-and-neck cancers is subject to inaccuracies. This study aims to evaluate contributions of magnetic resonance imaging (MRI), positron emission tomography (PET), and physical examination (PE) to GTV delineation in oropharyngeal cancer (OPC). Methods: Forty-one patients with OPC were studied. All underwent contrast-enhanced CT simulation scans (CECTs) that were registered with pretreatment PETs and MRIs. For each patient, three sets of primary and nodal GTV were contoured. First, reference GTVs (GTVref) were contoured by the treating radiation oncologist (RO) using CT, MRI, PET, and PE findings. Additional GTVs were created using fused CT/PET scans (GTVctpet) and CT/MRI scans (GTVctmr) by two other ROs blinded to GTVref. To compare GTVs, concordance indices (CI) were calculated by dividing the respective overlap volumes by overall volumes. To evaluate the contribution of PE, composite GTVs derived from CT, MRI, and PET (GTVctpetmr) were compared with GTVref. Results: For primary tumors, GTVref was significantly larger than GTVctpet and GTVctmr (p < 0.001). Although no significant difference in size was noted between GTVctpet and GTVctmr (p Z 0.39), there was poor concordance between them (CI Z 0.62). In addition, although CI (ctpetmr vs. ref) was low, it was significantly higher than CI (ctpet vs. ref) and CI (ctmr vs. ref) (p < 0.001), suggesting that neither modality should be used alone. Qualitative analyses to explain the low CI (ctpetmr vs. ref) revealed underestimation of mucosal disease when GTV was contoured without knowledge of PE findings. Similar trends were observed for nodal GTVs. However, CI (ctpet vs. ref), CI (ctmr vs. ref), and CI (ctpetmr vs. ref) were high (>0.75), indicating that although the modalities were complementary, the added benefit was small in the context of CECTs. In addition, PE did not aid greatly in nodal GTV delineation. Conclusion: PET and MRI are complementary and combined use is ideal. However, the low CI (ctpetmr vs. ref) particularly for primary tumors underscores the limitations of defining GTVs using imaging alone. PE is invaluable and must be incorporated. Ó 2012 Elsevier Inc. Keywords: GTV delineation, Head-and-neck cancer, Clinical examination, Contouring
Reprint requests to: Anuradha Thiagarajan, M.D., Department of Radiation Oncology, Memorial Sloan-Kettering Cancer Center, 1275
Int J Radiation Oncol Biol Phys, Vol. 83, No. 1, pp. 220e227, 2012 0360-3016/$ - see front matter Ó 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.ijrobp.2011.05.060
York Ave, New York, NY 10021. Tel: (212) 639-6800; Fax: (212) 7173104; E-mail: [email protected] Conflicts of interest: none.
Volume 83 Number 1 2012
Introduction In recent years, intensity-modulated radiation therapy (IMRT) has found widespread use in the management of head-and-neck cancers because of its ability to create sharp doseegradients between tumor volumes and neighboring critical structures. With the advent of such highly conformal radiation delivery techniques, precise delineation of target volumes has become critical. Even minor inaccuracies may lead to marginal misses of tumor or overdosage of surrounding normal tissues, results of which can be devastating. It is well documented that the sole utilization of computed tomography (CT) simulation scans in contouring gross tumor volume (GTV) is subject to a large degree of interobserver variability. In fact, in a study that evaluated the degree of concordance among eight experienced physicians who contoured identical GTVs based on contrast-enhanced CT scans alone, the overlap in contoured volumes was only 53% (1). Hence, attention has shifted to the incorporation of other imaging modalities such as magnetic resonance imaging (MRI) and more recently, positron emission tomography (PET), into the treatment planning process to clarify areas of tumor burden. MRIs offer several well recognized advantages over CTs. First, they provide superior softtissue contrast compared with CTs. Second, their multiplanar imaging capability permits better definition of the craniocaudal tumor extent. Third, although CTs are ideal at demonstrating cortical bone erosion, marrow infiltration is better appreciated on MRIs. Finally, MRIs are far less susceptible to image degradation caused by artifacts arising from dental amalgam (2e5). One recent study showed that the coregistration of MRI and planning CTs substantially improved interobserver variability in critical organ as well as target volume delineation, particularly in patients with intracranial tumor extension, heavy dental work, or contraindications to iodinated contrast agents (6). The integration of PET into the treatment planning process has become increasingly popular across a variety of tumor subsites including head and neck. There are several theorized advantages of using PET for target volume delineation. These include standardization of GTV delineation by minimization of interobserver variation; reduction in the size of the GTV on the basis of PET-derived metabolic information, facilitating sparing of normal tissues in the immediate vicinity; identification of regions of tumor extension either missed by or not readily apparent on CTand MRI; localization of PET-avid subvolumes of tumor to direct dose escalation; and pertreatment modification of target volumes to account for tumor shrinkage in adaptive planning strategies (7e10). Although the respective roles of PET and MRI in the target volume delineation of head-and-neck neoplasms are the subject of ongoing debate particularly in the context of contrast-enhanced CT simulation scans, it is undisputed that imaging plays a fundamental role in the initial workup of these patients and subsequently, in radiation treatment planning. What is considerably less appreciated is the role of careful physical examination in the assessment of disease extent in this anatomically complex subsite. This study aims to assess the relative contributions of MRI and PET to GTV delineation in patients with oropharyngeal cancer and evaluate the role, if any, of physical examination.
Methods and Materials This study was approved by the Memorial Sloan-Kettering Cancer Center institutional review board with a waiver of informed
Target volume delineation in oropharyngeal cancer
221
consent. Forty consecutive patients with oropharyngeal cancer treated at Memorial Sloan-Kettering Cancer Center were studied. Patient characteristics are shown on Table 1. All underwent contrast-enhanced CT simulation scans which were fused with pretreatment MRI and PET/CT scans obtained within 2 weeks of the date of simulation. All PET/CT scans were performed on a General Electric Discovery LS PET (Advance NXi)/CT (LightSpeed four-slice) scanner (GE HealthCare, Waukesha, WI) with an intrinsic resolution of 4.2-mm full-width at half maximum, using standardized image acquisition protocols that have been previously described (11). Fusion between the various datasets (CT/MRI and CT/PET) was performed using an in-house image registration software (ImgReg). This software uses a fully automated algorithm based on maximizing the mutual information in the respective image datasets, providing a rigid method of coregistration. The area over which fusion was performed was selected such that it centered over the area of interest (typically the GTV) and encompassed a wide area of soft tissue and bony structures as well as the periphery of the patient to provide as much information for the coregistration process as possible. For CT/MRI fusion, no distortion correction techniques were applied before coregistration as there is ample evidence to suggest that geometric inaccuracies arising from distortion from magnetic field inhomogeneities and nonlinearities in the MRI gradient are minimal in the head-and-neck region (4, 12). The quality of the registrations was subsequently evaluated subjectively using the brain, spinal cord, and mandibular bone as references, with corrections made in the axial, sagittal, coronal, or rotational directions if required. If the registration was deemed suboptimal for planning purposes even after manual fine-tuning because of variations in neck position, for example, the case was excluded from this study. For each patient, three sets of primary and nodal GTV were contoured. Reference “gold standard” GTVs (GTVref) were Table 1 Patient and tumor characteristics Age 55 >55 Gender Male Female Site Tonsil Base of tongue Pharyngeal wall T stage T1 T2 T3 T4 N stage N0 N1 N2 N3 Stage group I II III IV
18 (45%) 22 (55%) 31 (78%) 9 (22%) 22 (55%) 17 (43%) 1 (2%) 5 17 9 9
(13%) (43%) (22%) (22%)
3 9 26 2
(8%) (22%) (65%) (5%)
1 1 6 32
(2%) (2%) (15%) (81%)
222
Thiagarajan et al.
contoured by the treating radiation oncologist based on CT, MRI, and PET scans as well as findings on physical examination and nasoendoscopy. Additional GTV datasets were created using fused CT/PET scans (GTVctpet) and CT/MRI scans (GTVctmr) by two other radiation oncologists working in consensus. Prior to tumor delineation, they were provided with a brief clinical history and relevant radiologic reports but were blinded to GTVref and clinical examination findings. For each case, a minimum interval of 4 weeks was required between PET- and MRI-based target volume delineation to minimize bias. In addition, to maintain consistency in delineation of primary and nodal targets, the following guidelines were adhered to. For CTs, standard CT criteria for soft-tissue infiltration (anatomic distortion or asymmetry, contrast enhancement, and infiltration/obliteration of fat planes) and for cartilage/ bone invasion (presence of frank osteolysis, cortical erosion, abnormally increased density of normally hypodense cartilage, encasement of cartilaginous or bony structures by tumor) were used. Similarly, the guidelines used for the identification of nodal metastases on CT were as follows: nodal size (a cutoff of 10 mm was used for the shortest axial diameter of most cervical nodes with two notable exceptions: 15 mm for jugulodigastric nodes and 5 mm for retropharyngeal nodes), heterogeneous enhancement, ring enhancement, or presence of a central region of low attenuation indicative of necrosis, nodal shape (round vs. elongated), nodal grouping (i.e., presence of 3 lymph nodes, each with a minimal diameter of 8e10 mm), and extracapsular extension defined by reticulation of the fat adjacent to the external margin of a lymph node (13). For GTVctmr, the MRI sequence with the best tumor visualization was used for delineation. GTVs were typically delineated on the T2-weighted images. Pre- and postcontrast T1weighted images were used at the discretion of the observer. Soft-tissue asymmetry, presence of abnormal nodular or infiltrative tissue, fat replacement, and hypointensity of tissues were considered as indicative of malignancy on T2 images. MRI criteria for assessment of nodal metastases were identical to those used for CTs (13). Contrast-enhanced fat-suppressed T1-weighted MRIs were considered optimal sequences for evaluation of cervical nodal disease and were used, where available. For GTVctpet, visual interpretation of the PET signal was used for target definition using a standard contouring protocol with predefined window and color settings. All contours (GTVctmr and GTV ctpet) were subsequently verified by an experienced nuclear medicine physician and radiologist and appropriate modifications were made. After completion of this, absolute volumes for primary and nodal GTVref, GTVctpet, and GTVctmr were calculated and recorded separately. To compare GTVs derived from the various modalities, concordance indices (CI) were calculated by dividing
Fig. 1.
International Journal of Radiation Oncology Biology Physics the overlap volume by the overall volume for GTVctpet versus GTVctmr, GTVctpet versus GTVref and GTVctmr versus GTVref (Fig. 1). To evaluate the contribution of physical examination to GTV delineation, composite GTVs derived from CT, MRI, and PET (GTVctpetmr) were compared with GTVref. Pairwise t-tests were used for volumetric comparisons between the GTV datasets as well as for comparisons of concordance indices and statistical significance was confirmed via the performance of analysis of variance tests.
Results For primary tumors, mean GTVref was significantly larger than GTVctpet and GTVctmr (50.1 mL vs. 33.9 mL and 35.1 mL, respectively, p < 0.001). No significant difference in size was noted between GTVctpet and GTVctmr (p Z 0.39). However, there was poor concordance between GTVctpet and GTVctmr (CI Z 0.62). Of note, although CI (ctpetmr vs. ref) was low (0.62), it was significantly higher than CI (ctpet vs. ref) and CI (ctmr vs. ref) (0.54 and 0.55, respectively; p < 0.001) suggesting that neither modality should be used in isolation (Table 2). Qualitative analyses to explain the low CI (ctpetmr vs. ref) revealed that the superficial extent of primary disease was often underestimated when GTV was contoured without knowledge of physical examination findings. Similar trends were observed for nodal GTVs. Mean GTVref was larger than GTVctpet and GTVctmr but did not reach statistical significance. In addition, although CI (ctpetmr vs. ref) was higher than CI (ctpet vs. ref) and CI (ctmr vs. ref), the respective concordance indices were all greater than 0.75 and there was no statistically significant difference between them. This indicates that while the imaging modalities may be complementary to each other in the assessment of nodal GTV, the added benefit is small in the context of contrast-enhanced CT simulation scans. In addition, physical examination did not aid greatly in nodal GTV delineation as evidenced by a high CI (ctpetmr vs. ref) of 0.84.
Discussion In the era of high-precision radiotherapy, accurate tumor delineation is crucial. There is often significant variability between clinicians with regards to what constitutes GTV. Although this interobserver variation in target delineation has been noted in virtually all anatomic regions, the greatest discrepancies have been observed in the delineation of head-and-neck carcinomas
Derivation of concordance index by dividing overlap volume by overall volume. (A) Overlap volume. (B) Overall volume.
Volume 83 Number 1 2012 Table 2
Target volume delineation in oropharyngeal cancer
223
Volumetric comparisons and comparisons of concordance between GTV datasets
Primary tumor Nodes Primary tumor Nodes
GTVctpet mean volume (mL)
GTVctmr mean volume (mL)
GTVref mean volume (mL)
Statistics
33.9 34.9
34.9 34.4
50.1 40.9
F (2,117) Z 3.945, p Z 0.022 F (2,108) Z 0.307, p Z 0.73
CI (ctpet vs. ref)
CI (ctmr vs. ref)
CI (ctpetmr vs. ref)
Statistics
0.54 0.76
0.55 0.77
0.62 0.84
F (2,117) Z 7.597, p Z 0.001 F (2,108) Z 0.496, p Z 0.61
Abbreviations: CT Z computed tomography; GTV Z gross tumor volume; GTVctmr Z composite GTVs from CT/MRI scans; GTVctpet Z composite GTVs from CT/PET scans; GTVctpetmr Z composite GTVs from CT, MRI and PET; MRI Z magnetic resonance imaging; PET Z positron emission tomography; ref Z reference.
both with regard to the size of the target volumes and the relative location of the outlined volumes with respect to each other (14). These inaccuracies generate systematic errors that will be reproduced during each fraction and propagated throughout the course of radiation. Depending on the magnitude of the error, the clinical impact may be significant, leading to variations in tumor control and/or adverse events. Over the years, numerous recommendations have been made to improve consistency of target definition including continued education of radiation oncologists, collaboration with radiologists, nuclear medicine physicians, and, where relevant, surgeons, as well as implementation of detailed protocols for tumor delineation (15, 16). At present, the incorporation of multimodality imaging to increase the accuracy of GTV delineation in head-and-neck malignancies is generating intense interest. Radiation oncologists have traditionally used CTs for radiotherapy planning as it provides the spatial accuracy and electron density information required for doseecalculation algorithms. However, it may be suboptimal in the assessment of certain headand-neck cancers, particularly those located in the oral cavity and oropharynx for several reasons. First, there is significant image degradation that occurs as a consequence of scatter artifacts from dental amalgam (6). This is an almost universal problem in headand-neck cancer patients who suffer from poor dentition. Second, even with contrast enhancement, CTs have a limited capacity to
discriminate between tumor and surrounding glottic musculature. As evidenced in our study, MRI has a distinct advantage in this regard because of its superior soft-tissue contrast, reduced interference from dental-work artifact, as well as the ability to view tumors in sagittal and coronal planes, allowing better definition of primary tumor extent (Fig. 2, 3). However, its susceptibility to geometric distortions at the edge of the field of view and the lack of tissue density information precludes its use as the sole modality for treatment planning purposes (2). Although coregistration of CT with MRI provides important complementary information and overcomes some shortcomings of each individual modality, these structural imaging techniques have inherent limitations in their capacity to differentiate tumor from neighboring normal tissues. This has led to interest in the incorporation of metabolic imaging to improve the delineation of viable tumor tissue in the head-andneck region. The utility of PET in radiation treatment planning for lung cancer is undisputed, particularly in the setting of atelectasis and/ or postobstructive pneumonia (17). However, its role in head-andneck cancer is still under investigation. Much of the published literature has only demonstrated that utilization of PET results in differences in the magnitude of the target volumes (mostly smaller volumes) and a reduction in interobserver variation (7, 8, 10). However, it is important to realize that smaller contoured volumes and interindividual consistency in volume delineation do not
Fig. 2. (A) Image degradation on computed tomography simulation scan resulting from dental artifacts. (B) Magnetic resonance imaging permitting better definition of extent of right-sided oropharyngeal cancer from lack of interference from dental artifact.
224
Thiagarajan et al.
International Journal of Radiation Oncology Biology Physics
Fig. 3. (A) Barely perceptible left base of tongue tumor on computed tomography (CT) simulation scan. (B) Magnetic resonance imaging demonstrating superior soft-tissue contrast and better appreciation of left base of tongue malignancy (arrow heads) barely visible on CT.
necessarily equate to superior target volumes. In addition, one of the most controversial and as yet unresolved issues in applying PET/CT in radiation planning is the method used for edge definition. Precise identification of tumor boundaries is of particular importance in the head-and-neck region where high radiation doses are delivered to lesions in close proximity to radiosensitive neural and/or optic structures. However, the poor spatial resolution of PET (w6e8 mm) and the partial volume effects causing blurring of the edges make this challenging. In addition, changing the PET window level can alter target volumes by a factor of 2 or more with smaller lesions being particularly susceptible to threshold changes (7, 8). Various automated and semiautomated operator-independent segmentation methods (e.g., standardized uptake valuee and threshold-based contouring approaches, source-background algorithms) have been proposed in the literature in efforts to reduce interobserver variation in PET-based GTV delineation (11, 18). Although the concept of objective image segmentation appears attractive in theory, these methods are subject to potential pitfalls in that they cannot reliably distinguish between neoplastic processes and physiologic tracer uptake. In the absence of conclusive high-quality data establishing superiority of one method over another, it is difficult to imagine that any of the previously listed automated contouring methods or rigid mathematical models derived from a single crude parameter such as the standardized uptake value could supplant the analytic capabilities of the trained human brain in differentiating tumor from foci of physiological uptake. We believe that carefully designed contouring protocols using predefined display settings (windowing, color scale) in combination with input from a nuclear medicine physician with expertise in PET/CT interpretation can provide highly reproducible and reliable results. Hence, this was the approach that was favored in our study. One of the most commonly cited benefits of the utilization of PET in radiotherapy planning is its ability to guide nodal GTV delineation by inclusion of nodes that are equivocal on conventional imaging modalities (8, 9). With the advent of highresolution CT scanners and the use of intravenous contrast, this benefit may be less modest than previously thought. In our study, while PET aided the identification of diseased nodes that were not
apparent or missed by CT and/or MRI, there were several instances where cervical nodes demonstrated obvious radiographic evidence of necrosis but no tracer uptake (Fig. 4). In fact, the sensitivity of PET in a clinically node-negative neck is only in the range of 50e70% (19). In a prospective study performed at Memorial Sloan-Kettering Cancer Center, 31 patients with oral cavity cancer and no clinical or radiographic evidence of lymph node metastases underwent PET/CT before elective neck dissection (20). Microscopic nodal infiltration was noted in approximately 30% of patients who were PET-negative. Even worse findings were documented in a recent meta-analysis which sought to reevaluate the diagnostic performance of PET (21). Here, PET correctly identified only half of all clinically node-negative patients with histopathologic evidence of nodal involvement. Hence, if any therapeutic decision had been based solely on PET findings, a significant proportion of patients would have been undertreated with prophylactic radiation doses rather than tumoricidal doses. This was amply demonstrated in a recent report of periparotid recurrence after definitive IMRT for locally advanced head-and-neck cancer (22). When pretreatment imaging studies were retrospectively reviewed, in two of three cases, nonspecific nodules were observed on MRI either within the parotid gland or in the periparotid region but in both cases, no abnormal PET activity was noted. In their concluding statements, the authors urged caution in the interpretation of PET findings and warned against being falsely reassured by a negative PET, particularly if there was a high index of clinical suspicion. It is well recognized that persistent or recurrent nodal disease predicts for distant failure and impacts adversely on overall survival in head-and-neck cancers. Hence, adequate coverage of gross nodal disease is critical and excessive reliance on PET for nodal GTV delineation may lead to significant underestimations with potentially dire consequences. Although the value of PET in cervical nodal evaluation has been well documented, what is less appreciated is the complementary role it plays to CT and MRI in the assessment of primary tumors in the following settings: in radiographically and clinically occult tumors (e.g., base of tongue malignancies), to distinguish tumor from reactive changes and/or scarring from prior treatment,
Volume 83 Number 1 2012
Target volume delineation in oropharyngeal cancer
225
Fig. 4. (A) Large necrotic node seen on computed tomography. (B) T2-weighted magnetic resonance imaging demonstrating fluid density within the same cervical node in keeping with necrosis. (C) Positron emission tomography scan in the same patient showing no signal uptake in the diseased node. and, finally, where metal or motion artifacts impair tumor visualization (Fig. 5). However, this needs to be balanced against the fact that physiologic tracer uptake in the major salivary glands, tonsils, and uninvolved base of tongue may often confound PET findings and as was done in this study, close collaboration with an experienced nuclear medicine physician is of utmost importance. When regions of nonoverlap between GTVctpet and GTVctmr for primary tumor were quantified in our study, it was noted that 16% of GTVctpetmr was identified only on PET. Likewise, 18% was observed only on MRI. However, one of the major shortcomings of our study and one that is common to many other studies of this nature is the lack of histopathologic correlation. In the absence of surgical resection, the true tumor volume and shape is unknown. Hence, there is no gold standard against which volumetric comparisons can be made. In the absence of such information and/or clinical outcome data, superiority of one modality over another cannot be established. The minimal available literature is somewhat contradictory with one
study claiming that the surgical specimens corresponded most closely to the PET-based volumes, whereas the other concluded that GTVctpetmr most accurately represented the ground truth (23, 24). At present, using either modality to shrink volumes and exclude potentially diseased regions from radiation portals is fraught with danger and the most prudent approach would be to use the composite of the volumes (i.e., GTVctpetmr) in conjunction with physical examination findings. The other weakness of this study was that the majority of PET and MRI scans were acquired separately from CT simulation scans without a flat couchtop or customized immobilization devices to reproduce treatment conditions or laser beams to verify patient alignment. We acknowledge that the ideal situation would be simultaneous acquisition of all three imaging studies in an integrated simulation suite. However, many patients had undergone MRI and PET scans during their initial workup and repeat scans within a short period were not feasible primarily because of
Fig. 5. (A). Radiographically occult tumor not visible on computed tomography simulation scan (note significant interference from dental amalgam). (B). Highly 18F-Fluoro-deoxyglucose (FDG)-avid lesion noted in the right oropharynx in the same patient.
226
Thiagarajan et al.
International Journal of Radiation Oncology Biology Physics
Fig. 6. Respective contributions of magnetic resonance imaging (MRI), positron emission tomography (PET), and physical examination to gross tumor volume (GTV) delineation. Green line denotes GTV derived from CT and PET (i.e., GTVctpet). Blue line denotes GTV derived from computed tomography (CT) and MRI (i.e., GTVctmr). Yellow line denotes GTV derived from CT, PET, MRI, and physical examination (i.e., GTVref). Note that GTVctmr and GTVctpet both underestimate the mucosal extent of disease.
issues with reimbursement. To minimize the impact of any matching errors on subsequent volume delineation, the accuracy of the coregistration was checked in detail before contouring. If the quality of fusion was considered unsatisfactory, the patient was excluded from our study. In addition, the PET and MRIs that were used were acquired within 2 weeks of the CT simulation date to minimize the impact of tumor progression on volume delineation. That said, in the search for accuracy in target delineation, hardware coregistration with integrated CT/PET/MRI scanners with standardized image acquisition protocols are the way of the future. Finally, no discussion about target volume delineation is complete without consideration of the role played by physical examination. With advances in imaging technology, physical examination is often viewed by radiation oncologists (RO) as time-consuming and redundant. Often, at the time of referral, the typical head-and-neck cancer patient has already been scanned, scoped, diagnosed, and staged. Hence, there is a temptation for the busy RO to omit physical examination altogether and to base subsequent treatment decisions on imaging studies and the assessment of referring surgeons. However, the fallacy in relying on imaging alone to define GTVs is borne out by a study evaluating the performance of pretreatment CT, MRI, and 18FFluoro-deoxyglucose (FDG)-PET in patients with laryngeal and oropharyngeal cancers, a small proportion of whom subsequently underwent total laryngectomy (23). Detailed comparison of PET-, MRI-, and CT-derived GTVs with resected tumor specimens showed that no one modality was 100% accurate. Although these GTVs were significantly larger than the macroscopic extent of disease, all three modalities underestimated the superficial spread of disease. In our study, similar findings were observed (Fig. 6). In fact, more than one-third of the GTVref (for primary tumor) was attributable solely to physical examination findings! Thus, although the information derived from cross-sectional imaging is fundamental in the evaluation of clinically occult submucosal extension as well as deep soft-tissue infiltration in head-and-neck neoplasms, mucosal tumor extent is often better assessed by physical examination.
In this regard, the head-and-neck region has a unique advantage compared with many other sites. It is easily amenable to clinical examination by inspection, palpation, and fiberoptic endoscopy. It must be stated that the physical examination referred to here is not a perfunctory exercise. It should not just be about making an observation of an abnormality but refining that observation and meticulously documenting findings in a detailed manner, taking clinical photographs where relevant. For example, in the assessment of a tonsillar cancer, in addition to measuring the dimensions of the lesion, careful observation and palpation to determine the presence and extent of palatine or base of tongue involvement is key. Traditionally and rightly so, there has been much emphasis placed on the understanding of normal head-and-neck imaging anatomy to assist in image interpretation and target volume delineation in this anatomically complex region. It forms the core teaching curriculum of radiation oncology residency programs and it is frequently the subject of continuing medical education workshops for ROs in practice. However, there needs to be an equal focus on the mastery of clinical examination skills: the artful and gentle use of the tongue blade to avoid stimulating the gag reflex that once activated, could hinder meaningful visualization of the upper aerodigestive tract, the maneuvers to aid indirect mirror examination, and fiberoptic endoscopy just to name a few. The results of this study serve as a timely reminder that radiation oncologists are first and foremost clinicians not technicians and that radiation oncology should not be a profession practiced before the computer screen.
Conclusion If target delineation is inaccurate, highly conformal radiation techniques such as IMRT are rendered ineffective. Each imaging modality (CT, MRI, and PET) adds complementary information, and their combined use is recommended. However, sole reliance on volumetric imaging to identify targets for delineation often results in significant underestimation of the superficial component
Volume 83 Number 1 2012 of disease. To obtain the most comprehensive information regarding tumor extent presently achievable without a surgical procedure, thorough physical examination and meticulous documentation of findings are invaluable and must be incorporated as an essential step in precise GTV delineation. Additional histopathologic correlation studies as well as large multicenter prospective trials analyzing clinical outcomes are needed to validate this emerging paradigm in the radiation treatment planning of head-and-neck cancers. In addition, in the current era of exploding health care costs, it would be prudent to perform cost-effectiveness analyses prior to making recommendations about the implementation and use of these imaging modalities in the wider radiation oncology community.
References 1. Cooper JS, Mukherji SK, Toledano AY, et al. An evaluation of the variability of tumor-shape definition derived by experienced observers from CT images of supraglottic carcinomas (ACRIN protocol 6658). Int J Radiat Oncol Biol Phys 2007;67:972e975. 2. Toonkel LM, Soila K, Gilbert D, et al. MRI assisted treatment planning for radiation therapy of the head and neck. Magn Reson Imaging 1988;6:315e319. 3. Rasch C, Keus R, Pameijer FA, et al. The potential impact of CT-MRI matching on tumor volume delineation in advanced head and neck cancer. Int J Radiat Oncol Biol Phys 1997;39:841e848. 4. Emami B, Sethi A, Petruzzelli GJ. Influence of MRI on target volume delineation and IMRT planning in nasopharyngeal carcinoma. Int J Radiat Oncol Biol Phys 2003;57:481e488. 5. Chung NN, Ting LL, Hsu WC, et al. Impact of magnetic resonance imaging versus CT on nasopharyngeal carcinoma: Primary tumor target delineation for radiotherapy. Head Neck 2004;26:241e246. 6. O’Daniel JC, Rosenthal DI, Garden AS, et al. The effect of dental artifacts, contrast media, and experience on interobserver contouring variations in head and neck anatomy. Am J Clin Oncol 2007;30:191e198. 7. MacManus M, Nestle U, Rosenzweig KE, et al. Use of PET and PET/CT for radiation therapy planning: IAEA expert report 2006e2007. Radiother Oncol 2009;91:85e94. 8. Troost EG, Schinagl DA, Bussink J, et al. Innovations in radiotherapy planning of head and neck cancers: Role of PET. J Nucl Med 2010;51: 66e76. 9. Fletcher JW, Djulbegovic B, Soares HP, et al. Recommendations on the use of 18F-FDG PET in oncology. J Nucl Med 2008;49:480e508. 10. Podoloff DA, Ball DW, Ben-Josef E, et al. NCCN task force: Clinical utility of PET in a variety of tumor types. J Natl Compr Canc Netw 2009;7(Suppl 2):S1eS26.
Target volume delineation in oropharyngeal cancer
227
11. Greco C, Nehmeh SA, Schoder H, et al. Evaluation of different methods of 18F-FDG-PET target volume delineation in the radiotherapy of head and neck cancer. Am J Clin Oncol 2008;31: 439e445. 12. Ahmed M, Schmidt M, Sohaib A, et al. The value of magnetic resonance imaging in target volume delineation of base of tongue tumoursda study using flexible surface coils. Radiother Oncol 2010; 94:161e167. 13. Som PM. Detection of metastasis in cervical lymph nodes: CT and MR criteria and differential diagnosis. AJR Am J Roentgenol 1992;158: 961e969. 14. Weiss E, Hess CF. The impact of gross tumor volume (GTV) and clinical target volume (CTV) definition on the total accuracy in radiotherapy theoretical aspects and practical experiences. Strahlenther Onkol 2003;179:21e30. 15. Nuyts S. Defining the target for radiotherapy of head and neck cancer. Cancer Imaging 2007;7(Spec No A):S50eS55. 16. Rasch C, Steenbakkers R, van Herk M. Target definition in prostate, head, and neck. Semin Radiat Oncol 2005;15:136e145. 17. Greco C, Rosenzweig K, Cascini GL, et al. Current status of PET/CT for tumour volume definition in radiotherapy treatment planning for non-small cell lung cancer (NSCLC). Lung Cancer 2007;57: 125e134. 18. Schinagl DA, Vogel WV, Hoffmann AL, et al. Comparison of five segmentation tools for 18F-fluoro-deoxy-glucose-positron emission tomography-based target volume definition in head and neck cancer. Int J Radiat Oncol Biol Phys 2007;69:1282e1289. 19. Iyer NG, Clark JR, Singham S, et al. Role of pretreatment (18)FDGPET/CT in surgical decision-making for head and neck cancers. Head Neck 2010;32(9):1202e1208. 20. Schoder H, Carlson DL, Kraus DH, et al. 18F-FDG PET/CT for detecting nodal metastases in patients with oral cancer staged N0 by clinical examination and CT/MRI. J Nucl Med 2006;47: 755e762. 21. Kyzas PA, Evangelou E, Denaxa-Kyza D, et al. 18F-fluorodeoxyglucose positron emission tomography to evaluate cervical node metastases in patients with head and neck squamous cell carcinoma: A meta-analysis. J Natl Cancer Inst 2008;100:712e720. 22. Cannon DM, Lee NY. Recurrence in region of spared parotid gland after definitive intensity-modulated radiotherapy for head and neck cancer. Int J Radiat Oncol Biol Phys 2008;70:660e665. 23. Daisne JF, Duprez T, Weynand B, et al. Tumor volume in pharyngolaryngeal squamous cell carcinoma: Comparison at CT, MR imaging, and FDG PET and validation with surgical specimen. Radiology 2004;233:93e100. 24. Wong WL, Hussain K, Chevretton E, et al. Validation and clinical application of computer-combined computed tomography and positron emission tomography with 2-[18F]fluoro-2-deoxy-D-glucose head and neck images. Am J Surg 1996;172:628e632.
Radiation Oncology biology
physics
www.redjournal.org
Clinical Investigation: Head and Neck Cancer
Target Volume Delineation in Oropharyngeal Cancer: Impact of PET, MRI, and Physical Examination Anuradha Thiagarajan, M.D.,* Nicola Caria, M.D.,* Heiko Scho¨der, M.D.,y N. Gopalakrishna Iyer, M.D.,z Suzanne Wolden, M.D.,* Richard J. Wong, M.D.,z Eric Sherman, M.D.,x Matthew G. Fury, M.D.,x and Nancy Lee, M.D.* From the Departments of *Radiation Oncology, yRadiology, zHead and Neck Surgery, and xMedical Oncology, Memorial Sloan-Kettering Cancer Center, New York, NY Received Jan 12, 2011, and in revised form May 29, 2011. Accepted for publication May 31, 2011
Summary In the era of high precision radiotherapy, accurate target delineation is crucial. Sole utilization of computed tomography scans in gross tumor volume delineation for head and neck cancers is subject to significant interobserver variation. This paper demonstrates that magnetic resonance imaging and positron emission tomography add valuable complementary information, and that their combined use is recommended. In addition, it shows that thorough physical examination is invaluable in assessing superficial tumor extent in oropharyngeal malignancies, a dimension that is often missed or underestimated by imaging alone.
Introduction: Sole utilization of computed tomography (CT) scans in gross tumor volume (GTV) delineation for head-and-neck cancers is subject to inaccuracies. This study aims to evaluate contributions of magnetic resonance imaging (MRI), positron emission tomography (PET), and physical examination (PE) to GTV delineation in oropharyngeal cancer (OPC). Methods: Forty-one patients with OPC were studied. All underwent contrast-enhanced CT simulation scans (CECTs) that were registered with pretreatment PETs and MRIs. For each patient, three sets of primary and nodal GTV were contoured. First, reference GTVs (GTVref) were contoured by the treating radiation oncologist (RO) using CT, MRI, PET, and PE findings. Additional GTVs were created using fused CT/PET scans (GTVctpet) and CT/MRI scans (GTVctmr) by two other ROs blinded to GTVref. To compare GTVs, concordance indices (CI) were calculated by dividing the respective overlap volumes by overall volumes. To evaluate the contribution of PE, composite GTVs derived from CT, MRI, and PET (GTVctpetmr) were compared with GTVref. Results: For primary tumors, GTVref was significantly larger than GTVctpet and GTVctmr (p < 0.001). Although no significant difference in size was noted between GTVctpet and GTVctmr (p Z 0.39), there was poor concordance between them (CI Z 0.62). In addition, although CI (ctpetmr vs. ref) was low, it was significantly higher than CI (ctpet vs. ref) and CI (ctmr vs. ref) (p < 0.001), suggesting that neither modality should be used alone. Qualitative analyses to explain the low CI (ctpetmr vs. ref) revealed underestimation of mucosal disease when GTV was contoured without knowledge of PE findings. Similar trends were observed for nodal GTVs. However, CI (ctpet vs. ref), CI (ctmr vs. ref), and CI (ctpetmr vs. ref) were high (>0.75), indicating that although the modalities were complementary, the added benefit was small in the context of CECTs. In addition, PE did not aid greatly in nodal GTV delineation. Conclusion: PET and MRI are complementary and combined use is ideal. However, the low CI (ctpetmr vs. ref) particularly for primary tumors underscores the limitations of defining GTVs using imaging alone. PE is invaluable and must be incorporated. Ó 2012 Elsevier Inc. Keywords: GTV delineation, Head-and-neck cancer, Clinical examination, Contouring
Reprint requests to: Anuradha Thiagarajan, M.D., Department of Radiation Oncology, Memorial Sloan-Kettering Cancer Center, 1275
Int J Radiation Oncol Biol Phys, Vol. 83, No. 1, pp. 220e227, 2012 0360-3016/$ - see front matter Ó 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.ijrobp.2011.05.060
York Ave, New York, NY 10021. Tel: (212) 639-6800; Fax: (212) 7173104; E-mail: [email protected] Conflicts of interest: none.
Volume 83 Number 1 2012
Introduction In recent years, intensity-modulated radiation therapy (IMRT) has found widespread use in the management of head-and-neck cancers because of its ability to create sharp doseegradients between tumor volumes and neighboring critical structures. With the advent of such highly conformal radiation delivery techniques, precise delineation of target volumes has become critical. Even minor inaccuracies may lead to marginal misses of tumor or overdosage of surrounding normal tissues, results of which can be devastating. It is well documented that the sole utilization of computed tomography (CT) simulation scans in contouring gross tumor volume (GTV) is subject to a large degree of interobserver variability. In fact, in a study that evaluated the degree of concordance among eight experienced physicians who contoured identical GTVs based on contrast-enhanced CT scans alone, the overlap in contoured volumes was only 53% (1). Hence, attention has shifted to the incorporation of other imaging modalities such as magnetic resonance imaging (MRI) and more recently, positron emission tomography (PET), into the treatment planning process to clarify areas of tumor burden. MRIs offer several well recognized advantages over CTs. First, they provide superior softtissue contrast compared with CTs. Second, their multiplanar imaging capability permits better definition of the craniocaudal tumor extent. Third, although CTs are ideal at demonstrating cortical bone erosion, marrow infiltration is better appreciated on MRIs. Finally, MRIs are far less susceptible to image degradation caused by artifacts arising from dental amalgam (2e5). One recent study showed that the coregistration of MRI and planning CTs substantially improved interobserver variability in critical organ as well as target volume delineation, particularly in patients with intracranial tumor extension, heavy dental work, or contraindications to iodinated contrast agents (6). The integration of PET into the treatment planning process has become increasingly popular across a variety of tumor subsites including head and neck. There are several theorized advantages of using PET for target volume delineation. These include standardization of GTV delineation by minimization of interobserver variation; reduction in the size of the GTV on the basis of PET-derived metabolic information, facilitating sparing of normal tissues in the immediate vicinity; identification of regions of tumor extension either missed by or not readily apparent on CTand MRI; localization of PET-avid subvolumes of tumor to direct dose escalation; and pertreatment modification of target volumes to account for tumor shrinkage in adaptive planning strategies (7e10). Although the respective roles of PET and MRI in the target volume delineation of head-and-neck neoplasms are the subject of ongoing debate particularly in the context of contrast-enhanced CT simulation scans, it is undisputed that imaging plays a fundamental role in the initial workup of these patients and subsequently, in radiation treatment planning. What is considerably less appreciated is the role of careful physical examination in the assessment of disease extent in this anatomically complex subsite. This study aims to assess the relative contributions of MRI and PET to GTV delineation in patients with oropharyngeal cancer and evaluate the role, if any, of physical examination.
Methods and Materials This study was approved by the Memorial Sloan-Kettering Cancer Center institutional review board with a waiver of informed
Target volume delineation in oropharyngeal cancer
221
consent. Forty consecutive patients with oropharyngeal cancer treated at Memorial Sloan-Kettering Cancer Center were studied. Patient characteristics are shown on Table 1. All underwent contrast-enhanced CT simulation scans which were fused with pretreatment MRI and PET/CT scans obtained within 2 weeks of the date of simulation. All PET/CT scans were performed on a General Electric Discovery LS PET (Advance NXi)/CT (LightSpeed four-slice) scanner (GE HealthCare, Waukesha, WI) with an intrinsic resolution of 4.2-mm full-width at half maximum, using standardized image acquisition protocols that have been previously described (11). Fusion between the various datasets (CT/MRI and CT/PET) was performed using an in-house image registration software (ImgReg). This software uses a fully automated algorithm based on maximizing the mutual information in the respective image datasets, providing a rigid method of coregistration. The area over which fusion was performed was selected such that it centered over the area of interest (typically the GTV) and encompassed a wide area of soft tissue and bony structures as well as the periphery of the patient to provide as much information for the coregistration process as possible. For CT/MRI fusion, no distortion correction techniques were applied before coregistration as there is ample evidence to suggest that geometric inaccuracies arising from distortion from magnetic field inhomogeneities and nonlinearities in the MRI gradient are minimal in the head-and-neck region (4, 12). The quality of the registrations was subsequently evaluated subjectively using the brain, spinal cord, and mandibular bone as references, with corrections made in the axial, sagittal, coronal, or rotational directions if required. If the registration was deemed suboptimal for planning purposes even after manual fine-tuning because of variations in neck position, for example, the case was excluded from this study. For each patient, three sets of primary and nodal GTV were contoured. Reference “gold standard” GTVs (GTVref) were Table 1 Patient and tumor characteristics Age 55 >55 Gender Male Female Site Tonsil Base of tongue Pharyngeal wall T stage T1 T2 T3 T4 N stage N0 N1 N2 N3 Stage group I II III IV
18 (45%) 22 (55%) 31 (78%) 9 (22%) 22 (55%) 17 (43%) 1 (2%) 5 17 9 9
(13%) (43%) (22%) (22%)
3 9 26 2
(8%) (22%) (65%) (5%)
1 1 6 32
(2%) (2%) (15%) (81%)
222
Thiagarajan et al.
contoured by the treating radiation oncologist based on CT, MRI, and PET scans as well as findings on physical examination and nasoendoscopy. Additional GTV datasets were created using fused CT/PET scans (GTVctpet) and CT/MRI scans (GTVctmr) by two other radiation oncologists working in consensus. Prior to tumor delineation, they were provided with a brief clinical history and relevant radiologic reports but were blinded to GTVref and clinical examination findings. For each case, a minimum interval of 4 weeks was required between PET- and MRI-based target volume delineation to minimize bias. In addition, to maintain consistency in delineation of primary and nodal targets, the following guidelines were adhered to. For CTs, standard CT criteria for soft-tissue infiltration (anatomic distortion or asymmetry, contrast enhancement, and infiltration/obliteration of fat planes) and for cartilage/ bone invasion (presence of frank osteolysis, cortical erosion, abnormally increased density of normally hypodense cartilage, encasement of cartilaginous or bony structures by tumor) were used. Similarly, the guidelines used for the identification of nodal metastases on CT were as follows: nodal size (a cutoff of 10 mm was used for the shortest axial diameter of most cervical nodes with two notable exceptions: 15 mm for jugulodigastric nodes and 5 mm for retropharyngeal nodes), heterogeneous enhancement, ring enhancement, or presence of a central region of low attenuation indicative of necrosis, nodal shape (round vs. elongated), nodal grouping (i.e., presence of 3 lymph nodes, each with a minimal diameter of 8e10 mm), and extracapsular extension defined by reticulation of the fat adjacent to the external margin of a lymph node (13). For GTVctmr, the MRI sequence with the best tumor visualization was used for delineation. GTVs were typically delineated on the T2-weighted images. Pre- and postcontrast T1weighted images were used at the discretion of the observer. Soft-tissue asymmetry, presence of abnormal nodular or infiltrative tissue, fat replacement, and hypointensity of tissues were considered as indicative of malignancy on T2 images. MRI criteria for assessment of nodal metastases were identical to those used for CTs (13). Contrast-enhanced fat-suppressed T1-weighted MRIs were considered optimal sequences for evaluation of cervical nodal disease and were used, where available. For GTVctpet, visual interpretation of the PET signal was used for target definition using a standard contouring protocol with predefined window and color settings. All contours (GTVctmr and GTV ctpet) were subsequently verified by an experienced nuclear medicine physician and radiologist and appropriate modifications were made. After completion of this, absolute volumes for primary and nodal GTVref, GTVctpet, and GTVctmr were calculated and recorded separately. To compare GTVs derived from the various modalities, concordance indices (CI) were calculated by dividing
Fig. 1.
International Journal of Radiation Oncology Biology Physics the overlap volume by the overall volume for GTVctpet versus GTVctmr, GTVctpet versus GTVref and GTVctmr versus GTVref (Fig. 1). To evaluate the contribution of physical examination to GTV delineation, composite GTVs derived from CT, MRI, and PET (GTVctpetmr) were compared with GTVref. Pairwise t-tests were used for volumetric comparisons between the GTV datasets as well as for comparisons of concordance indices and statistical significance was confirmed via the performance of analysis of variance tests.
Results For primary tumors, mean GTVref was significantly larger than GTVctpet and GTVctmr (50.1 mL vs. 33.9 mL and 35.1 mL, respectively, p < 0.001). No significant difference in size was noted between GTVctpet and GTVctmr (p Z 0.39). However, there was poor concordance between GTVctpet and GTVctmr (CI Z 0.62). Of note, although CI (ctpetmr vs. ref) was low (0.62), it was significantly higher than CI (ctpet vs. ref) and CI (ctmr vs. ref) (0.54 and 0.55, respectively; p < 0.001) suggesting that neither modality should be used in isolation (Table 2). Qualitative analyses to explain the low CI (ctpetmr vs. ref) revealed that the superficial extent of primary disease was often underestimated when GTV was contoured without knowledge of physical examination findings. Similar trends were observed for nodal GTVs. Mean GTVref was larger than GTVctpet and GTVctmr but did not reach statistical significance. In addition, although CI (ctpetmr vs. ref) was higher than CI (ctpet vs. ref) and CI (ctmr vs. ref), the respective concordance indices were all greater than 0.75 and there was no statistically significant difference between them. This indicates that while the imaging modalities may be complementary to each other in the assessment of nodal GTV, the added benefit is small in the context of contrast-enhanced CT simulation scans. In addition, physical examination did not aid greatly in nodal GTV delineation as evidenced by a high CI (ctpetmr vs. ref) of 0.84.
Discussion In the era of high-precision radiotherapy, accurate tumor delineation is crucial. There is often significant variability between clinicians with regards to what constitutes GTV. Although this interobserver variation in target delineation has been noted in virtually all anatomic regions, the greatest discrepancies have been observed in the delineation of head-and-neck carcinomas
Derivation of concordance index by dividing overlap volume by overall volume. (A) Overlap volume. (B) Overall volume.
Volume 83 Number 1 2012 Table 2
Target volume delineation in oropharyngeal cancer
223
Volumetric comparisons and comparisons of concordance between GTV datasets
Primary tumor Nodes Primary tumor Nodes
GTVctpet mean volume (mL)
GTVctmr mean volume (mL)
GTVref mean volume (mL)
Statistics
33.9 34.9
34.9 34.4
50.1 40.9
F (2,117) Z 3.945, p Z 0.022 F (2,108) Z 0.307, p Z 0.73
CI (ctpet vs. ref)
CI (ctmr vs. ref)
CI (ctpetmr vs. ref)
Statistics
0.54 0.76
0.55 0.77
0.62 0.84
F (2,117) Z 7.597, p Z 0.001 F (2,108) Z 0.496, p Z 0.61
Abbreviations: CT Z computed tomography; GTV Z gross tumor volume; GTVctmr Z composite GTVs from CT/MRI scans; GTVctpet Z composite GTVs from CT/PET scans; GTVctpetmr Z composite GTVs from CT, MRI and PET; MRI Z magnetic resonance imaging; PET Z positron emission tomography; ref Z reference.
both with regard to the size of the target volumes and the relative location of the outlined volumes with respect to each other (14). These inaccuracies generate systematic errors that will be reproduced during each fraction and propagated throughout the course of radiation. Depending on the magnitude of the error, the clinical impact may be significant, leading to variations in tumor control and/or adverse events. Over the years, numerous recommendations have been made to improve consistency of target definition including continued education of radiation oncologists, collaboration with radiologists, nuclear medicine physicians, and, where relevant, surgeons, as well as implementation of detailed protocols for tumor delineation (15, 16). At present, the incorporation of multimodality imaging to increase the accuracy of GTV delineation in head-and-neck malignancies is generating intense interest. Radiation oncologists have traditionally used CTs for radiotherapy planning as it provides the spatial accuracy and electron density information required for doseecalculation algorithms. However, it may be suboptimal in the assessment of certain headand-neck cancers, particularly those located in the oral cavity and oropharynx for several reasons. First, there is significant image degradation that occurs as a consequence of scatter artifacts from dental amalgam (6). This is an almost universal problem in headand-neck cancer patients who suffer from poor dentition. Second, even with contrast enhancement, CTs have a limited capacity to
discriminate between tumor and surrounding glottic musculature. As evidenced in our study, MRI has a distinct advantage in this regard because of its superior soft-tissue contrast, reduced interference from dental-work artifact, as well as the ability to view tumors in sagittal and coronal planes, allowing better definition of primary tumor extent (Fig. 2, 3). However, its susceptibility to geometric distortions at the edge of the field of view and the lack of tissue density information precludes its use as the sole modality for treatment planning purposes (2). Although coregistration of CT with MRI provides important complementary information and overcomes some shortcomings of each individual modality, these structural imaging techniques have inherent limitations in their capacity to differentiate tumor from neighboring normal tissues. This has led to interest in the incorporation of metabolic imaging to improve the delineation of viable tumor tissue in the head-andneck region. The utility of PET in radiation treatment planning for lung cancer is undisputed, particularly in the setting of atelectasis and/ or postobstructive pneumonia (17). However, its role in head-andneck cancer is still under investigation. Much of the published literature has only demonstrated that utilization of PET results in differences in the magnitude of the target volumes (mostly smaller volumes) and a reduction in interobserver variation (7, 8, 10). However, it is important to realize that smaller contoured volumes and interindividual consistency in volume delineation do not
Fig. 2. (A) Image degradation on computed tomography simulation scan resulting from dental artifacts. (B) Magnetic resonance imaging permitting better definition of extent of right-sided oropharyngeal cancer from lack of interference from dental artifact.
224
Thiagarajan et al.
International Journal of Radiation Oncology Biology Physics
Fig. 3. (A) Barely perceptible left base of tongue tumor on computed tomography (CT) simulation scan. (B) Magnetic resonance imaging demonstrating superior soft-tissue contrast and better appreciation of left base of tongue malignancy (arrow heads) barely visible on CT.
necessarily equate to superior target volumes. In addition, one of the most controversial and as yet unresolved issues in applying PET/CT in radiation planning is the method used for edge definition. Precise identification of tumor boundaries is of particular importance in the head-and-neck region where high radiation doses are delivered to lesions in close proximity to radiosensitive neural and/or optic structures. However, the poor spatial resolution of PET (w6e8 mm) and the partial volume effects causing blurring of the edges make this challenging. In addition, changing the PET window level can alter target volumes by a factor of 2 or more with smaller lesions being particularly susceptible to threshold changes (7, 8). Various automated and semiautomated operator-independent segmentation methods (e.g., standardized uptake valuee and threshold-based contouring approaches, source-background algorithms) have been proposed in the literature in efforts to reduce interobserver variation in PET-based GTV delineation (11, 18). Although the concept of objective image segmentation appears attractive in theory, these methods are subject to potential pitfalls in that they cannot reliably distinguish between neoplastic processes and physiologic tracer uptake. In the absence of conclusive high-quality data establishing superiority of one method over another, it is difficult to imagine that any of the previously listed automated contouring methods or rigid mathematical models derived from a single crude parameter such as the standardized uptake value could supplant the analytic capabilities of the trained human brain in differentiating tumor from foci of physiological uptake. We believe that carefully designed contouring protocols using predefined display settings (windowing, color scale) in combination with input from a nuclear medicine physician with expertise in PET/CT interpretation can provide highly reproducible and reliable results. Hence, this was the approach that was favored in our study. One of the most commonly cited benefits of the utilization of PET in radiotherapy planning is its ability to guide nodal GTV delineation by inclusion of nodes that are equivocal on conventional imaging modalities (8, 9). With the advent of highresolution CT scanners and the use of intravenous contrast, this benefit may be less modest than previously thought. In our study, while PET aided the identification of diseased nodes that were not
apparent or missed by CT and/or MRI, there were several instances where cervical nodes demonstrated obvious radiographic evidence of necrosis but no tracer uptake (Fig. 4). In fact, the sensitivity of PET in a clinically node-negative neck is only in the range of 50e70% (19). In a prospective study performed at Memorial Sloan-Kettering Cancer Center, 31 patients with oral cavity cancer and no clinical or radiographic evidence of lymph node metastases underwent PET/CT before elective neck dissection (20). Microscopic nodal infiltration was noted in approximately 30% of patients who were PET-negative. Even worse findings were documented in a recent meta-analysis which sought to reevaluate the diagnostic performance of PET (21). Here, PET correctly identified only half of all clinically node-negative patients with histopathologic evidence of nodal involvement. Hence, if any therapeutic decision had been based solely on PET findings, a significant proportion of patients would have been undertreated with prophylactic radiation doses rather than tumoricidal doses. This was amply demonstrated in a recent report of periparotid recurrence after definitive IMRT for locally advanced head-and-neck cancer (22). When pretreatment imaging studies were retrospectively reviewed, in two of three cases, nonspecific nodules were observed on MRI either within the parotid gland or in the periparotid region but in both cases, no abnormal PET activity was noted. In their concluding statements, the authors urged caution in the interpretation of PET findings and warned against being falsely reassured by a negative PET, particularly if there was a high index of clinical suspicion. It is well recognized that persistent or recurrent nodal disease predicts for distant failure and impacts adversely on overall survival in head-and-neck cancers. Hence, adequate coverage of gross nodal disease is critical and excessive reliance on PET for nodal GTV delineation may lead to significant underestimations with potentially dire consequences. Although the value of PET in cervical nodal evaluation has been well documented, what is less appreciated is the complementary role it plays to CT and MRI in the assessment of primary tumors in the following settings: in radiographically and clinically occult tumors (e.g., base of tongue malignancies), to distinguish tumor from reactive changes and/or scarring from prior treatment,
Volume 83 Number 1 2012
Target volume delineation in oropharyngeal cancer
225
Fig. 4. (A) Large necrotic node seen on computed tomography. (B) T2-weighted magnetic resonance imaging demonstrating fluid density within the same cervical node in keeping with necrosis. (C) Positron emission tomography scan in the same patient showing no signal uptake in the diseased node. and, finally, where metal or motion artifacts impair tumor visualization (Fig. 5). However, this needs to be balanced against the fact that physiologic tracer uptake in the major salivary glands, tonsils, and uninvolved base of tongue may often confound PET findings and as was done in this study, close collaboration with an experienced nuclear medicine physician is of utmost importance. When regions of nonoverlap between GTVctpet and GTVctmr for primary tumor were quantified in our study, it was noted that 16% of GTVctpetmr was identified only on PET. Likewise, 18% was observed only on MRI. However, one of the major shortcomings of our study and one that is common to many other studies of this nature is the lack of histopathologic correlation. In the absence of surgical resection, the true tumor volume and shape is unknown. Hence, there is no gold standard against which volumetric comparisons can be made. In the absence of such information and/or clinical outcome data, superiority of one modality over another cannot be established. The minimal available literature is somewhat contradictory with one
study claiming that the surgical specimens corresponded most closely to the PET-based volumes, whereas the other concluded that GTVctpetmr most accurately represented the ground truth (23, 24). At present, using either modality to shrink volumes and exclude potentially diseased regions from radiation portals is fraught with danger and the most prudent approach would be to use the composite of the volumes (i.e., GTVctpetmr) in conjunction with physical examination findings. The other weakness of this study was that the majority of PET and MRI scans were acquired separately from CT simulation scans without a flat couchtop or customized immobilization devices to reproduce treatment conditions or laser beams to verify patient alignment. We acknowledge that the ideal situation would be simultaneous acquisition of all three imaging studies in an integrated simulation suite. However, many patients had undergone MRI and PET scans during their initial workup and repeat scans within a short period were not feasible primarily because of
Fig. 5. (A). Radiographically occult tumor not visible on computed tomography simulation scan (note significant interference from dental amalgam). (B). Highly 18F-Fluoro-deoxyglucose (FDG)-avid lesion noted in the right oropharynx in the same patient.
226
Thiagarajan et al.
International Journal of Radiation Oncology Biology Physics
Fig. 6. Respective contributions of magnetic resonance imaging (MRI), positron emission tomography (PET), and physical examination to gross tumor volume (GTV) delineation. Green line denotes GTV derived from CT and PET (i.e., GTVctpet). Blue line denotes GTV derived from computed tomography (CT) and MRI (i.e., GTVctmr). Yellow line denotes GTV derived from CT, PET, MRI, and physical examination (i.e., GTVref). Note that GTVctmr and GTVctpet both underestimate the mucosal extent of disease.
issues with reimbursement. To minimize the impact of any matching errors on subsequent volume delineation, the accuracy of the coregistration was checked in detail before contouring. If the quality of fusion was considered unsatisfactory, the patient was excluded from our study. In addition, the PET and MRIs that were used were acquired within 2 weeks of the CT simulation date to minimize the impact of tumor progression on volume delineation. That said, in the search for accuracy in target delineation, hardware coregistration with integrated CT/PET/MRI scanners with standardized image acquisition protocols are the way of the future. Finally, no discussion about target volume delineation is complete without consideration of the role played by physical examination. With advances in imaging technology, physical examination is often viewed by radiation oncologists (RO) as time-consuming and redundant. Often, at the time of referral, the typical head-and-neck cancer patient has already been scanned, scoped, diagnosed, and staged. Hence, there is a temptation for the busy RO to omit physical examination altogether and to base subsequent treatment decisions on imaging studies and the assessment of referring surgeons. However, the fallacy in relying on imaging alone to define GTVs is borne out by a study evaluating the performance of pretreatment CT, MRI, and 18FFluoro-deoxyglucose (FDG)-PET in patients with laryngeal and oropharyngeal cancers, a small proportion of whom subsequently underwent total laryngectomy (23). Detailed comparison of PET-, MRI-, and CT-derived GTVs with resected tumor specimens showed that no one modality was 100% accurate. Although these GTVs were significantly larger than the macroscopic extent of disease, all three modalities underestimated the superficial spread of disease. In our study, similar findings were observed (Fig. 6). In fact, more than one-third of the GTVref (for primary tumor) was attributable solely to physical examination findings! Thus, although the information derived from cross-sectional imaging is fundamental in the evaluation of clinically occult submucosal extension as well as deep soft-tissue infiltration in head-and-neck neoplasms, mucosal tumor extent is often better assessed by physical examination.
In this regard, the head-and-neck region has a unique advantage compared with many other sites. It is easily amenable to clinical examination by inspection, palpation, and fiberoptic endoscopy. It must be stated that the physical examination referred to here is not a perfunctory exercise. It should not just be about making an observation of an abnormality but refining that observation and meticulously documenting findings in a detailed manner, taking clinical photographs where relevant. For example, in the assessment of a tonsillar cancer, in addition to measuring the dimensions of the lesion, careful observation and palpation to determine the presence and extent of palatine or base of tongue involvement is key. Traditionally and rightly so, there has been much emphasis placed on the understanding of normal head-and-neck imaging anatomy to assist in image interpretation and target volume delineation in this anatomically complex region. It forms the core teaching curriculum of radiation oncology residency programs and it is frequently the subject of continuing medical education workshops for ROs in practice. However, there needs to be an equal focus on the mastery of clinical examination skills: the artful and gentle use of the tongue blade to avoid stimulating the gag reflex that once activated, could hinder meaningful visualization of the upper aerodigestive tract, the maneuvers to aid indirect mirror examination, and fiberoptic endoscopy just to name a few. The results of this study serve as a timely reminder that radiation oncologists are first and foremost clinicians not technicians and that radiation oncology should not be a profession practiced before the computer screen.
Conclusion If target delineation is inaccurate, highly conformal radiation techniques such as IMRT are rendered ineffective. Each imaging modality (CT, MRI, and PET) adds complementary information, and their combined use is recommended. However, sole reliance on volumetric imaging to identify targets for delineation often results in significant underestimation of the superficial component
Volume 83 Number 1 2012 of disease. To obtain the most comprehensive information regarding tumor extent presently achievable without a surgical procedure, thorough physical examination and meticulous documentation of findings are invaluable and must be incorporated as an essential step in precise GTV delineation. Additional histopathologic correlation studies as well as large multicenter prospective trials analyzing clinical outcomes are needed to validate this emerging paradigm in the radiation treatment planning of head-and-neck cancers. In addition, in the current era of exploding health care costs, it would be prudent to perform cost-effectiveness analyses prior to making recommendations about the implementation and use of these imaging modalities in the wider radiation oncology community.
References 1. Cooper JS, Mukherji SK, Toledano AY, et al. An evaluation of the variability of tumor-shape definition derived by experienced observers from CT images of supraglottic carcinomas (ACRIN protocol 6658). Int J Radiat Oncol Biol Phys 2007;67:972e975. 2. Toonkel LM, Soila K, Gilbert D, et al. MRI assisted treatment planning for radiation therapy of the head and neck. Magn Reson Imaging 1988;6:315e319. 3. Rasch C, Keus R, Pameijer FA, et al. The potential impact of CT-MRI matching on tumor volume delineation in advanced head and neck cancer. Int J Radiat Oncol Biol Phys 1997;39:841e848. 4. Emami B, Sethi A, Petruzzelli GJ. Influence of MRI on target volume delineation and IMRT planning in nasopharyngeal carcinoma. Int J Radiat Oncol Biol Phys 2003;57:481e488. 5. Chung NN, Ting LL, Hsu WC, et al. Impact of magnetic resonance imaging versus CT on nasopharyngeal carcinoma: Primary tumor target delineation for radiotherapy. Head Neck 2004;26:241e246. 6. O’Daniel JC, Rosenthal DI, Garden AS, et al. The effect of dental artifacts, contrast media, and experience on interobserver contouring variations in head and neck anatomy. Am J Clin Oncol 2007;30:191e198. 7. MacManus M, Nestle U, Rosenzweig KE, et al. Use of PET and PET/CT for radiation therapy planning: IAEA expert report 2006e2007. Radiother Oncol 2009;91:85e94. 8. Troost EG, Schinagl DA, Bussink J, et al. Innovations in radiotherapy planning of head and neck cancers: Role of PET. J Nucl Med 2010;51: 66e76. 9. Fletcher JW, Djulbegovic B, Soares HP, et al. Recommendations on the use of 18F-FDG PET in oncology. J Nucl Med 2008;49:480e508. 10. Podoloff DA, Ball DW, Ben-Josef E, et al. NCCN task force: Clinical utility of PET in a variety of tumor types. J Natl Compr Canc Netw 2009;7(Suppl 2):S1eS26.
Target volume delineation in oropharyngeal cancer
227
11. Greco C, Nehmeh SA, Schoder H, et al. Evaluation of different methods of 18F-FDG-PET target volume delineation in the radiotherapy of head and neck cancer. Am J Clin Oncol 2008;31: 439e445. 12. Ahmed M, Schmidt M, Sohaib A, et al. The value of magnetic resonance imaging in target volume delineation of base of tongue tumoursda study using flexible surface coils. Radiother Oncol 2010; 94:161e167. 13. Som PM. Detection of metastasis in cervical lymph nodes: CT and MR criteria and differential diagnosis. AJR Am J Roentgenol 1992;158: 961e969. 14. Weiss E, Hess CF. The impact of gross tumor volume (GTV) and clinical target volume (CTV) definition on the total accuracy in radiotherapy theoretical aspects and practical experiences. Strahlenther Onkol 2003;179:21e30. 15. Nuyts S. Defining the target for radiotherapy of head and neck cancer. Cancer Imaging 2007;7(Spec No A):S50eS55. 16. Rasch C, Steenbakkers R, van Herk M. Target definition in prostate, head, and neck. Semin Radiat Oncol 2005;15:136e145. 17. Greco C, Rosenzweig K, Cascini GL, et al. Current status of PET/CT for tumour volume definition in radiotherapy treatment planning for non-small cell lung cancer (NSCLC). Lung Cancer 2007;57: 125e134. 18. Schinagl DA, Vogel WV, Hoffmann AL, et al. Comparison of five segmentation tools for 18F-fluoro-deoxy-glucose-positron emission tomography-based target volume definition in head and neck cancer. Int J Radiat Oncol Biol Phys 2007;69:1282e1289. 19. Iyer NG, Clark JR, Singham S, et al. Role of pretreatment (18)FDGPET/CT in surgical decision-making for head and neck cancers. Head Neck 2010;32(9):1202e1208. 20. Schoder H, Carlson DL, Kraus DH, et al. 18F-FDG PET/CT for detecting nodal metastases in patients with oral cancer staged N0 by clinical examination and CT/MRI. J Nucl Med 2006;47: 755e762. 21. Kyzas PA, Evangelou E, Denaxa-Kyza D, et al. 18F-fluorodeoxyglucose positron emission tomography to evaluate cervical node metastases in patients with head and neck squamous cell carcinoma: A meta-analysis. J Natl Cancer Inst 2008;100:712e720. 22. Cannon DM, Lee NY. Recurrence in region of spared parotid gland after definitive intensity-modulated radiotherapy for head and neck cancer. Int J Radiat Oncol Biol Phys 2008;70:660e665. 23. Daisne JF, Duprez T, Weynand B, et al. Tumor volume in pharyngolaryngeal squamous cell carcinoma: Comparison at CT, MR imaging, and FDG PET and validation with surgical specimen. Radiology 2004;233:93e100. 24. Wong WL, Hussain K, Chevretton E, et al. Validation and clinical application of computer-combined computed tomography and positron emission tomography with 2-[18F]fluoro-2-deoxy-D-glucose head and neck images. Am J Surg 1996;172:628e632.