- Email: [email protected]
On-Line Localization of the Lumpectomy Cavity Using Surgical Clips
Int. J. Radiation Oncology Biol. Phys., Vol. 69, No. 4, pp. 1305–1309, 2007 Copyright Ó 2007 Elsevier Inc. Printed in the USA. All rights reserved 0360-3016/07/$–see front matter
doi:10.1016/j.ijrobp.2007.07.2365
PHYSICS CONTRIBUTION
ON-LINE LOCALIZATION OF THE LUMPECTOMY CAVITY USING SURGICAL CLIPS LEONARD H. KIM, M.S., A.MUS.D.,* JOHN WONG, PH.D.,y AND DI YAN, D.SC.* * Department of Radiation Oncology, William Beaumont Hospital, Royal Oak, MI; and y Department of Radiation Oncology and Molecular Radiation Sciences, School of Medicine, Johns Hopkins University, Baltimore, MD Purpose: To present an on-line image guidance procedure for external beam accelerated partial breast irradiation based on cone-beam computed tomography (CBCT) imaging of surgical clips; and to estimate the possible clinical target volume (CTV) to planning target volume (PTV) margin reduction allowed by this technique. Methods and Materials: Clips in the CBCT image are detected automatically using in-house software. The treatment couch is translated according to the shift in the clips’ center of mass between the planning and CBCT images. Three components for the PTV margin are considered: (1) breathing, (2) surrogate error (i.e., error in cavity position after perfect setup to clips), and (3) residual error (i.e., error arising from the inability to execute a perfect setup to clips due to technological limitations, such as couch travel precision). These factors were input into a standard formula for CTV-to-PTV margin calculation. Results: The average magnitude of clip-based corrections was 7 ± 2 mm (10 patients, 44 fractions). After localization, the residual error magnitude was 1.6 ± 1.3 mm, justifying an isotropic CTV-to-PTV margin of approximately 6 mm, including breathing and surrogate error. Conclusions: On-line localization of the lumpectomy cavity using surgical clips is technically feasible from the standpoint of equipment, time, and process, making possible a decreased CTV-to-PTV margin for accelerated partial breast irradiation. Because the procedure is exclusively target based, additional monitoring of critical structures may be advisable. Ó 2007 Elsevier Inc. Partial breast irradiation, Image guidance, Cone-beam CT.
Several options exist for the setup of patients undergoing external beam accelerated partial breast irradiation (APBI). These include setup to (1) stable landmarks, whether external (tattoos) or internal (bony registration), (2) the breast, whether as a volume or as a surface, or (3) the lumpectomy cavity, as defined by surgical clips or seroma. Common to all of these approaches is the use of a surrogate to represent the location of the clinical target volume (CTV). The margin added to the CTV to generate a planning target volume (PTV) is traditionally interpreted as an accommodation for setup error and breathing. One way of approaching the setup error component of the PTV margin is to view it as a combination of surrogate error and residual error. Surrogate error is the error in target position after perfect setup to the surrogate. Residual error is the error associated with imperfect setup to the surrogate, such as due to user or technological imprecision.
Measuring the effectiveness of surrogates for the lumpectomy cavity has been the subject of a number of recent abstracts. Langen et al. (1) compare manual registrations based on breast surface and rib cage to a baseline registration of the seroma. Gierga et al. (2) use surgical clips placed on the periphery of the lumpectomy cavity as the baseline ‘‘truth’’ against which setups based on laser alignment to skin marks, chest wall alignment, and surface imaging are evaluated. Hasan et al. (3) test the ability of manual bony and surface registrations, as well as calculated corrections based on surgical clips, to track shifts in the cavity center of mass (COM). The use of surgical clips or seroma as baselines in the aforementioned studies suggests they are viewed as the best surrogates for the CTV. Surgical clips have the advantage of relatively better radiographic contrast. To be sure, though, the use of surgical clips poses problems, such as
Reprint requests to: Leonard Kim, M.S., Department of Radiation Oncology, William Beaumont Hospital, 3601 West Thirteen Mile Road, Royal Oak, MI 48073. Tel: (248) 551-7037; Fax: (248) 551-3784; E-mail: [email protected] Supported in part by Elekta. Presented in part at poster discussion sessions at the 2004 and 2005 Annual Meetings of the American Society for Therapeutic Radiology and Oncology (ASTRO) (October 3–7, 2004, Atlanta, GA; October 16–20, 2005, Denver, CO). Conflict of interest: none.
Acknowledgments—Radim Cernej wrote the three-dimensional clip localization software. Cone-beam CT images were often acquired by our therapists. One of them, Jennifer Wloch, eventually took much of the responsibility for identifying potential study patients and doing image registration. Dr. Carlos Vargas initially assisted with patient recruiting. Daniel Letourneau, Doug Drake, and Geoff Hugo provided technical expertise on Synergy. Drs. Frank Vicini and Alvaro Martinez provided general clinical support. Received Dec 11, 2006, and in revised form July 23, 2007. Accepted for publication July 24, 2007.
INTRODUCTION
1305
1306
I. J. Radiation Oncology d Biology d Physics
migration and sampling (i.e., a relatively small number of points is taken to represent a whole). A feasibility study by Weed et al. (4) concludes that there is a mean surrogate error of approximately 3 mm when surgical clips are used to localize the cavity COM. From this, they predict that clip-based, on-line setup correction would allow a reduction in CTVto-PTV margin from 10 mm to 5 mm. Clinically, they proposed using radiographs acquired at near-orthogonal angles for clip imaging and localization. A software tool was developed to manage this process (5), but difficulties with marker visualization rendered the procedure unfeasible in some test cases and uncertain to the point of limited utility in others (6). The present study had two primary purposes: (1) to present our experience with on-line breast patient setup correction using cone-beam CT (CBCT) imaging of surgical clips followed by couch translation, and (2) to estimate an appropriate CTV-to-PTV margin when this on-line correction procedure is used in conjunction with APBI. METHODS AND MATERIALS On-line setup correction procedure Ten whole-breast radiotherapy patients with surgical clips in the lumpectomy cavity were evaluated for this study. The study procedure involved (1) initial patient setup using laser alignment to skin marks, (2) initial CBCT acquisition to assess initial setup error, (3) couch translation based on the initial CBCT, and (4) a second CBCT to assess residual error. Each patient underwent this procedure once per week over the course of their treatment, giving 4 to 6 sessions per patient and 49 sessions total. Five patients were imaged using a research version of the Elekta Synergy with XVI software version 3.1 (Elekta, Norcross, GA). For these patients, both couch corrections and clip positions were determined on line using in-house software described below. Midway into the study the system was upgraded to the clinically released version of Synergy and XVI version 3.5. The remaining 5 patients were imaged with this system. For these patients the Elekta Synergy XVI software was used to generate the on-line couch correction, and the in-house software was later used to determine clip positions off line. Only the inhouse software results for setup and residual error were used in this study. The CBCT technique adopted using the clinical version of Synergy was small field of view, 120 kVp, 0.8 mAs per projection (80 mA, 10 ms) with a 200 scan range. On the basis of in-house dose measurements with a phantom, this technique is estimated to give an approximately 2.3-cGy skin dose per scan. The complete procedure took approximately 15 min, including deployment of CBCT components, scanning, reconstruction, localization, and couch correction. The postcorrection CBCT scan is included in this time estimate as well. The in-house software automatically detects clips in the CBCT image by self-adjusting a voxel value threshold until an appropriate number of objects are captured. Software running time is approximately 20 s. A coordinate system is established by assigning the center of the CBCT volume the same coordinates as the planning isocenter. The coordinates of each clip in the CBCT are then defined by the location of its brightest voxel. A COM is calculated from the average of all of the individual clips’ coordinates. To obtain a couch correction, this COM is subtracted from a reference COM, obtained by averaging the coordinates of each clip contour’s centroid in the planning CT. The difference between the CBCT and reference COMs is returned to the user as a couch translation.
Volume 69, Number 4, 2007
PTV margin calculation Three components were considered for the PTV margin: residual error, surrogate error, and breathing. Residual error was measured by comparing the intended couch shift to the actual shift in the clips COM between pre- and postcorrection CBCTs. Surrogate error values were re-evaluated using a subgroup (12 patients) of the 28 patients used in Weed et al. (4). The 12 patients were chosen for having imaging schedules compatible with treatment under the Radiation Therapy Oncology Group (RTOG) 0413/National Surgical Adjuvant Breast and Bowel Project (NSABP) B-39 protocol (i.e., planning CT within 6 weeks of surgery and a localization CT within 31 days of planning) (7). Because only a single surrogate error measurement was available for each patient in the Weed et al. study, it could not be determined from these data to what extent surrogate error is systematic or random. For the purposes of this study, surrogate error was considered a systematic error. For each patient, a mean setup error (m) and standard deviation (s) were computed. From these data, two parameters were calculated: the standard deviation of mean setup errors, S(mi) and the root mean square of standard deviations, RMS(si) (8). Surrogate error was added in quadrature to S(mi) (i.e., it was considered an additional systematic error). A breathing component of 1.5 mm was added in quadrature to RMS(si) in each orthogonal direction. This value was chosen after consultation of the existing literature (4, 9–13) as well as our own fluoroscopic images of 8 whole-breast radiotherapy patients with clips, 5 of whom were also among our CBCT study patients. The postcorrection PTV margin was calculated as qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi P 2:33 ð ðmi Þ2 þRMSðsi Þ2 Þ. Because the population standard deviaqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi P P tion, p ; z ð ðmi Þ2 þRMSðsi Þ2 Þ, this margin is roughly equivalent to that obtained using the two-parameter model at a 98% confidence level (8).
RESULTS Individual patient mean setup errors (m) and standard deviations (s) are shown in Fig. 1. The precorrection mean and standard deviation of the individual mean setup errors, M(mi) S(mi), were 2.7 4.3, 1.4 3.1, and 2.5 2.5 mm in the right–left (RL), anterior–posterior (AP), and superior–inferior (SI) directions, respectively. The root mean square and standard deviation of the individual standard deviations, RMS(si) S(si), before correction were 2.2 1.4, 1.7 0.4, and 1.6 0.7 mm in the RL, AP, and SI directions, respectively. After correction, the mean and standard deviation of the individual mean residual errors, M(mi) S(mi), were 0.2 0.9, 0.4 1.4, and 0.0 1.1 mm in the RL, AP, and SI directions, respectively. The root mean square and standard deviation of the individual standard deviations, RMS(si) S(si), after correction were 0.6 0.2, 1.0 0.5, and 0.7 0.3 mm in the RL, AP, and SI directions, respectively. The global mean and standard deviation of surrogate errors were calculated to be 0.4 2.4, 0.6 1.7, and 0.3 2.3 mm for RL, AP, and SI directions, respectively, for the 12 patients from the Weed et al. study with imaging dates consistent with the NSABP B-39/RTOG 0413 protocol. The CTV-to-PTV margin allowable when clip-based, online setup correction is used was calculated to be 6 mm in each orthogonal direction.
Lumpectomy cavity localization using clips d L. H. KIM et al.
Fig. 1. Patient-specific setup errors, before and after correction (mean and 1 standard deviation). Ant = anterior; Post = posterior; Sup = superior; Inf = inferior.
DISCUSSION Setup correction procedure The study patients who underwent this procedure received whole-breast irradiation from two tangent beams. In that situation it was easy to set up the patient and couch to allow both treatment and a 200 CBCT scan without any collision issue. In more recent experience with actual image-guided APBI treatments, we found it strongly advisable, before the beginning of treatment, to assess whether such a setup is possible (i.e., a setup whereby all scans, portal images, and treatment geometries can be executed without having to move the couch or patient because of clearance issues). We found that the lateral and longitudinal couch positions may need to be narrowly specified to achieve this. An alternative solution is to scan the patient in an out-of-treatment position that physically allows the CBCT scan while ensuring that the target and other relevant structures remain in the field of view for registration purposes. Registration is performed to determine the ‘‘true’’ treatment position and the necessary couch ‘‘correction’’ required to reach it. The CBCT components are then retracted (withdrawing the couch if necessary to do this) before the couch is moved to the calculated treatment position. We have found this method to work well, with one drawback being that a postcorrection verification CBCT cannot be acquired in actual treatment position. PTV margin calculation Our precorrection setup data suggest that, without setup correction, margins of 12, 9, and 9 mm are required for 98% confidence in the RL, AP, and SI directions, respectively, if M(mi) = 0. Although we obtained values for M(mi)
1307
of 2.7, 1.4, and 2.5 mm in the RL, AP, and SI directions, respectively, our data are not inconsistent with M(mi) = 0 because of the magnitude of S(si) and our small patient sample, with the possible exception of SI. If M(mi) is really non-zero, this indicates a miscalibration between simulation and treatment setups that should be corrected before application of the quoted margins. Assuming M(mi) = 0, the 10-mm CTVto-PTV margin called for by the NSABP B-39/RTOG 0413 protocol (7) seems appropriate, with the possible exception of LR, for which the observed errors suggest that a 10-mm margin offers only 94% confidence. For this direction a margin of 12 mm or more might be considered to attain 98% confidence. This analysis assumes that there is negligible difference in precorrection setup error statistics of whole-breast patients, who made up our study population, and APBI patients. At our institution, at least, this seems reasonable because the patients are tattooed and set up identically. Further support for this assumption comes from Gierga et al. (2), who measured setup error after laser-based setup of APBI patients and obtained errors of the same magnitude as ours. According to our calculations, the use of our setup correction procedure would allow a CTV-to-PTV margin of 6 mm, a 4-mm reduction of the margin specified in NSABP B-39/ RTOG 0413. As noted above, the protocol value may even be low in one direction. Baglan et al. (10) obtained the 10mm figure partly on the basis of on an anticipated setup error of 5 mm. Our results and those of Gierga et al. suggest that this may have been slightly optimistic. On the other hand, Baglan et al. allowed 5 mm of margin for breathing motion, which is more than what we allow. It must be noted that estimates of breathing motion for APBI patients vary widely in the literature, ranging from 1 to 10 mm full excursion (4, 9– 13). Baglan et al.’s measurements are among the highest in the literature. Examination of their data suggests that their method of manually registering three CT scans to obtain range-of-motion estimates could have contributed to the larger values. Nevertheless, there is disagreement in the literature regarding magnitude of breathing motion, and its resolution could affect our margin calculation. Regarding margin calculation, we noted above that the patient data from the Weed et al. study do not allow determination of whether surrogate error is systematic or random. When one considers possible causes of surrogate error—contouring, clip migration, and sampling (i.e., clips are a small, nonrandom sample of the cavity)—it seems plausible that surrogate error for an individual patient is largely systematic. Ongoing work in which 6 APBI patients to date have received multiple CT scans of clips and cavity indicates that this is probably true. In that case, it may be possible to account for surrogate error largely through correction rather than the PTV margin. For example, a CT scan acquired just before the start of a patient’s treatment course could be used with the planning CT to estimate the magnitude and direction of that patient’s surrogate error. That can be addressed with an additional, global couch correction for every treatment fraction. The PTV margin would then need only include random variation of
I. J. Radiation Oncology d Biology d Physics
1308
Volume 69, Number 4, 2007
Table 1. Comparison with results from White et al.* Precorrection
Postcorrection
Parameter
LR
AP
SI
LR
AP
SI
White systematic Kim S(mi)
2.7 4.3
1.7 3.1
2.4 2.5
0.8 0.9
0.7 1.4
0.8 1.1
2.4 2.2 (1.4)
2.2 1.7 (0.4)
2.9 1.6 (0.7)
1.5 0.6 (0.2)
1.5 1.0 (0.5)
1.6 0.7 (0.3)
8.8 11.3
8.8 8.2
8.8 6.8
3.6 2.4
3.6 4.0
3.6 3.1
White random Kim RMS(si) (s(si)) White margin Kim margin (setup only)
Abbreviations: LR = left–right; AP = anterior–posterior; SI = superior–inferior. Values are millimeters. * Reference 14.
surrogate error—for the 6 APBI patients mentioned above, the RMS(si) of surrogate errors was approximately 0.8 mm. It should be emphasized that the just-described procedure is at present speculative and untested. A recent study by White et al. (14) describes CBCT-based, on-line APBI setup correction using manual registration of the chest wall and skin. A comparison of their results with ours is given in Table 1. For this comparison, surrogate error and breathing were removed from our margin estimates because these were not included in White et al. The two studies seem generally consistent despite the use of different surrogate structures and localization methods. This suggests that global setup error is presently a relatively greater issue than internal target motion for treatment accuracy. This hypothesis is supported by the observation that couch corrections based on clips vs. those based on whole-breast registration are well correlated (Fig. 2; r = 0.8) with a mean difference between a clip-based correction and a whole-breast-based correction in a given direction of 2.2 1.5 mm. These data
Fig. 2. Correlation between clip-based and whole-breast-based couch corrections.
are based on 5 of our study patients (24 images) for whom both types of registration were performed. The agreement in post-correction measurements in the two studies supports our assumption that the use of whole-breast patients rather than APBI patients does not affect the measurement of residual error. One of the key assumptions in this study is that the cavity COM is a suitable surrogate for the CTV. The CTV is normally generated by applying a margin to the lumpectomy cavity as visualized in the planning CT. Numerous studies have shown that the lumpectomy cavity’s shape and volume can change significantly over time (4, 15–20). Even within the comparatively short time frames of APBI, large changes have been observed. For the 12 patients in the Weed et al. study with imaging dates consistent with the RTOG 0413/ NSABP B-39 protocol, the mean volume reduction was 45% between planning and localization CTs. In recent work at our institution we observed a volume reduction of 30% 35% between planning and treatment in 16 APBI patients. In addition, volume changes were observed to be generally non-isotropic.
Fig. 3. Example of nonuniform cavity shrinkage.
Lumpectomy cavity localization using clips d L. H. KIM et al.
It is not clear how and to what extent the CTVs change with the cavity. If the CTV is essentially static, regardless of cavity volume change, using surgical clips for setup correction would be inadvisable because surgical clips track the cavity. On the other hand, if CTVs do change with the cavity over time, the nonuniformity of this change still raises issues given that the cavity COM is moving because of local volume changes, not target translation. For example, Fig. 3 shows cavity contours from two CTs acquired 3 weeks apart. The shrinkage is clearly non-uniform— although the smaller contour is bounded by the original, the center of the cavity has moved 12 mm between the two CTs. Yet it may be preferable to leave the setup as is and not make a cavity COMbased couch correction of 12 mm in the direction of the lung. If volume changes result in large relative shifts between the cavity and other structures, then global localization using
1309
the whole breast, surface, or chest wall might be preferable to cavity localization.
CONCLUSIONS On-line localization of the lumpectomy cavity using surgical clips is technically feasible from the standpoint of equipment, time, and process. Using this procedure, PTV margins for external beam APBI can be reduced to approximately 6 mm. Further reduction would require better performance and characterization of the target surrogate. Because this image guidance procedure is exclusively target based, additional monitoring of organs at risk may be advisable. Alternatively, localization based on the whole breast and chest wall could be considered.
REFERENCES 1. Langen K, Buchholz D, Victor G, et al. Evaluation of breast tumor bed alignment techniques for image guided radiotherapy [abstract]. Int J Radiat Oncol Biol Phys 2005;63(Suppl. 1): S179. 2. Gierga DP, Turcotte JC, Riboldi M, et al. Comparison of target registration errors for multiple modalities in image-guided partial breast irradiation [abstract]. Int J Radiat Oncol Biol Phys 2006;66(Suppl. 1):S103–S104. 3. Hasan Y, Kim L, Weed D, et al. Image guidance in external beam accelerated partial breast irradiation: Comparison of surrogates for the lumpectomy cavity [abstract]. Int J Radiat Oncol Biol Phys 2005;63(Suppl. 1):S138. 4. Weed DW, Yan D, Martinez AA, et al. The validity of surgical clips as a radiographic surrogate for the lumpectomy cavity in image-guided accelerated partial breast irradiation. Int J Radiat Oncol Biol Phys 2004;60:484–492. 5. Wong J, Watt L, Yan D, et al. ARTIST: Work flow for image guided radiation therapy. In: Yi BY, Ahn SD, Choi EK, et al., editors. Proceedings of the XIVth International Conference on the Use of Computers in Radiation Therapy; 2004 May 1013; Seoul, South Korea. Seoul: Jeong; 2004. p. 417–420. 6. Kim L, Cernej R, Vargas C, et al. Reduction of PTV margin for accelerated partial breast irradiation using on-line detection of surgical clips [virtual poster]. 2004 ASTRO Annual Meeting Abstracts and Virtual Posters [web site]. Fairfax, VA: American Society for Therapeutic Radiology and Oncology; 2004. Available at: http://astro.abstractsnet.com/handouts/000132_1133post. pdf. Accessed June 5, 2006. 7. National Surgical Adjuvant Breast and Bowel Project, Radiation Therapy Oncology Group. A randomized phase III study of conventional whole breast irradiation (WBI) versus partial breast irradiation (PBI) for women with stage 0, I, or II breast cancer. Philadelphia: Radiation Therapy Oncology Group; 2006. Available from: http://www.rtog.org/members/protocols/ 0413/0413.pdf. Accessed October 24, 2006. 8. Yan D, Lockman D, Martinez A, et al. Computed tomography guided management of interfractional patient variation. Semin Radiat Oncol 2005;15:168–179. 9. Bert C, Metheany KG, Doppke KP, et al. Clinical experience with a 3D surface patient setup system for alignment of partial-breast irradiation patients. Int J Radiat Oncol Biol Phys 2006;64:1265–1274. 10. Baglan KL, Sharpe MB, Jaffray D, et al. Accelerated partial breast irradiation using 3D conformal radiation therapy (3DCRT). Int J Radiat Oncol Biol Phys 2003;55:302–311.
11. Lee R, Suh H, Moon B, et al. Evaluation of breast motion as assessed by breast clips during normal breathing. In: Yi BY, Ahn SD, Choi EK, et al., editors. Proceedings of the XIVth International Conference on the Use of Computers in Radiation Therapy; 2004 May 10-13; Seoul, South Korea. Seoul: Jeong; 2004. p. 442–444. 12. Chopra S, Dinshaw KA, Kamble R, et al. Breast movement during normal and deep breathing, respiratory training and set up errors: Implications for external beam partial breast irradiation. Br J Radiol 2006;79:766–773. 13. Yue NJ, Li X, Beriwal S, et al. The intrafraction motion induced dosimetric impacts in breast 3D radiation treatment: A 4DCT based study. Med Phys 2007;34:2789–2800. 14. White EA, Cho J, Vallis KA, et al. Cone beam computed tomography guidance for setup of patients receiving accelerated partial breast irradiation. Int J Radiat Oncol Biol Phys 2007; 68:547–554. 15. Goldberg H, Prosnitz R, Olson J, et al. Definition of postlumpectomy tumor bed for radiotherapy boost field planning: CT versus surgical clips. Int J Radiat Oncol Biol Phys 2005;63: 209–213. 16. Oh KS, Kong F, Griffith KA, et al. Planning the breast tumor bed boost: Changes in the excision cavity volume and surgical scar location after breast-conserving surgery and whole-breast irradiation. Int J Radiat Oncol Biol Phys 2006;66:680–686. 17. Jacobson G, Betts V, Smith B. Change in volume of lumpectomy cavity during external-beam irradiation of the intact breast. Int J Radiat Oncol Biol Phys 2006;65:1161–1164. 18. Kubicek G, Kimler B, Bell M, et al. Changes in breast cavity contours after lumpectomy [virtual poster]. 2006 ASTRO Annual Meeting Abstracts and Virtual Posters [web site]. Fairfax, VA: American Society for Therapeutic Radiology and Oncology; 2006. Available from: http://astro2006.abstractsnet.com/ aoriginal.wcs?entryid=000060. Accessed February 20, 2007. 19. Vicini F, Remouchamps V, Wallace M, et al. Ongoing clinical experience utilizing 3D conformal external beam radiotherapy to deliver partial-breast irradiation in patients with early-stage breast cancer treated with breast-conserving therapy. Int J Radiat Oncol Biol Phys 2003;57:1247–1253. 20. Kader H, Truong P, Panades M, et al. Is breast seroma the appropriate guide for PTV determination in accelerated partial breast irradiation planning? [Abstract]. Radiother Oncol 2004; 72(Suppl.):S6.
doi:10.1016/j.ijrobp.2007.07.2365
PHYSICS CONTRIBUTION
ON-LINE LOCALIZATION OF THE LUMPECTOMY CAVITY USING SURGICAL CLIPS LEONARD H. KIM, M.S., A.MUS.D.,* JOHN WONG, PH.D.,y AND DI YAN, D.SC.* * Department of Radiation Oncology, William Beaumont Hospital, Royal Oak, MI; and y Department of Radiation Oncology and Molecular Radiation Sciences, School of Medicine, Johns Hopkins University, Baltimore, MD Purpose: To present an on-line image guidance procedure for external beam accelerated partial breast irradiation based on cone-beam computed tomography (CBCT) imaging of surgical clips; and to estimate the possible clinical target volume (CTV) to planning target volume (PTV) margin reduction allowed by this technique. Methods and Materials: Clips in the CBCT image are detected automatically using in-house software. The treatment couch is translated according to the shift in the clips’ center of mass between the planning and CBCT images. Three components for the PTV margin are considered: (1) breathing, (2) surrogate error (i.e., error in cavity position after perfect setup to clips), and (3) residual error (i.e., error arising from the inability to execute a perfect setup to clips due to technological limitations, such as couch travel precision). These factors were input into a standard formula for CTV-to-PTV margin calculation. Results: The average magnitude of clip-based corrections was 7 ± 2 mm (10 patients, 44 fractions). After localization, the residual error magnitude was 1.6 ± 1.3 mm, justifying an isotropic CTV-to-PTV margin of approximately 6 mm, including breathing and surrogate error. Conclusions: On-line localization of the lumpectomy cavity using surgical clips is technically feasible from the standpoint of equipment, time, and process, making possible a decreased CTV-to-PTV margin for accelerated partial breast irradiation. Because the procedure is exclusively target based, additional monitoring of critical structures may be advisable. Ó 2007 Elsevier Inc. Partial breast irradiation, Image guidance, Cone-beam CT.
Several options exist for the setup of patients undergoing external beam accelerated partial breast irradiation (APBI). These include setup to (1) stable landmarks, whether external (tattoos) or internal (bony registration), (2) the breast, whether as a volume or as a surface, or (3) the lumpectomy cavity, as defined by surgical clips or seroma. Common to all of these approaches is the use of a surrogate to represent the location of the clinical target volume (CTV). The margin added to the CTV to generate a planning target volume (PTV) is traditionally interpreted as an accommodation for setup error and breathing. One way of approaching the setup error component of the PTV margin is to view it as a combination of surrogate error and residual error. Surrogate error is the error in target position after perfect setup to the surrogate. Residual error is the error associated with imperfect setup to the surrogate, such as due to user or technological imprecision.
Measuring the effectiveness of surrogates for the lumpectomy cavity has been the subject of a number of recent abstracts. Langen et al. (1) compare manual registrations based on breast surface and rib cage to a baseline registration of the seroma. Gierga et al. (2) use surgical clips placed on the periphery of the lumpectomy cavity as the baseline ‘‘truth’’ against which setups based on laser alignment to skin marks, chest wall alignment, and surface imaging are evaluated. Hasan et al. (3) test the ability of manual bony and surface registrations, as well as calculated corrections based on surgical clips, to track shifts in the cavity center of mass (COM). The use of surgical clips or seroma as baselines in the aforementioned studies suggests they are viewed as the best surrogates for the CTV. Surgical clips have the advantage of relatively better radiographic contrast. To be sure, though, the use of surgical clips poses problems, such as
Reprint requests to: Leonard Kim, M.S., Department of Radiation Oncology, William Beaumont Hospital, 3601 West Thirteen Mile Road, Royal Oak, MI 48073. Tel: (248) 551-7037; Fax: (248) 551-3784; E-mail: [email protected] Supported in part by Elekta. Presented in part at poster discussion sessions at the 2004 and 2005 Annual Meetings of the American Society for Therapeutic Radiology and Oncology (ASTRO) (October 3–7, 2004, Atlanta, GA; October 16–20, 2005, Denver, CO). Conflict of interest: none.
Acknowledgments—Radim Cernej wrote the three-dimensional clip localization software. Cone-beam CT images were often acquired by our therapists. One of them, Jennifer Wloch, eventually took much of the responsibility for identifying potential study patients and doing image registration. Dr. Carlos Vargas initially assisted with patient recruiting. Daniel Letourneau, Doug Drake, and Geoff Hugo provided technical expertise on Synergy. Drs. Frank Vicini and Alvaro Martinez provided general clinical support. Received Dec 11, 2006, and in revised form July 23, 2007. Accepted for publication July 24, 2007.
INTRODUCTION
1305
1306
I. J. Radiation Oncology d Biology d Physics
migration and sampling (i.e., a relatively small number of points is taken to represent a whole). A feasibility study by Weed et al. (4) concludes that there is a mean surrogate error of approximately 3 mm when surgical clips are used to localize the cavity COM. From this, they predict that clip-based, on-line setup correction would allow a reduction in CTVto-PTV margin from 10 mm to 5 mm. Clinically, they proposed using radiographs acquired at near-orthogonal angles for clip imaging and localization. A software tool was developed to manage this process (5), but difficulties with marker visualization rendered the procedure unfeasible in some test cases and uncertain to the point of limited utility in others (6). The present study had two primary purposes: (1) to present our experience with on-line breast patient setup correction using cone-beam CT (CBCT) imaging of surgical clips followed by couch translation, and (2) to estimate an appropriate CTV-to-PTV margin when this on-line correction procedure is used in conjunction with APBI. METHODS AND MATERIALS On-line setup correction procedure Ten whole-breast radiotherapy patients with surgical clips in the lumpectomy cavity were evaluated for this study. The study procedure involved (1) initial patient setup using laser alignment to skin marks, (2) initial CBCT acquisition to assess initial setup error, (3) couch translation based on the initial CBCT, and (4) a second CBCT to assess residual error. Each patient underwent this procedure once per week over the course of their treatment, giving 4 to 6 sessions per patient and 49 sessions total. Five patients were imaged using a research version of the Elekta Synergy with XVI software version 3.1 (Elekta, Norcross, GA). For these patients, both couch corrections and clip positions were determined on line using in-house software described below. Midway into the study the system was upgraded to the clinically released version of Synergy and XVI version 3.5. The remaining 5 patients were imaged with this system. For these patients the Elekta Synergy XVI software was used to generate the on-line couch correction, and the in-house software was later used to determine clip positions off line. Only the inhouse software results for setup and residual error were used in this study. The CBCT technique adopted using the clinical version of Synergy was small field of view, 120 kVp, 0.8 mAs per projection (80 mA, 10 ms) with a 200 scan range. On the basis of in-house dose measurements with a phantom, this technique is estimated to give an approximately 2.3-cGy skin dose per scan. The complete procedure took approximately 15 min, including deployment of CBCT components, scanning, reconstruction, localization, and couch correction. The postcorrection CBCT scan is included in this time estimate as well. The in-house software automatically detects clips in the CBCT image by self-adjusting a voxel value threshold until an appropriate number of objects are captured. Software running time is approximately 20 s. A coordinate system is established by assigning the center of the CBCT volume the same coordinates as the planning isocenter. The coordinates of each clip in the CBCT are then defined by the location of its brightest voxel. A COM is calculated from the average of all of the individual clips’ coordinates. To obtain a couch correction, this COM is subtracted from a reference COM, obtained by averaging the coordinates of each clip contour’s centroid in the planning CT. The difference between the CBCT and reference COMs is returned to the user as a couch translation.
Volume 69, Number 4, 2007
PTV margin calculation Three components were considered for the PTV margin: residual error, surrogate error, and breathing. Residual error was measured by comparing the intended couch shift to the actual shift in the clips COM between pre- and postcorrection CBCTs. Surrogate error values were re-evaluated using a subgroup (12 patients) of the 28 patients used in Weed et al. (4). The 12 patients were chosen for having imaging schedules compatible with treatment under the Radiation Therapy Oncology Group (RTOG) 0413/National Surgical Adjuvant Breast and Bowel Project (NSABP) B-39 protocol (i.e., planning CT within 6 weeks of surgery and a localization CT within 31 days of planning) (7). Because only a single surrogate error measurement was available for each patient in the Weed et al. study, it could not be determined from these data to what extent surrogate error is systematic or random. For the purposes of this study, surrogate error was considered a systematic error. For each patient, a mean setup error (m) and standard deviation (s) were computed. From these data, two parameters were calculated: the standard deviation of mean setup errors, S(mi) and the root mean square of standard deviations, RMS(si) (8). Surrogate error was added in quadrature to S(mi) (i.e., it was considered an additional systematic error). A breathing component of 1.5 mm was added in quadrature to RMS(si) in each orthogonal direction. This value was chosen after consultation of the existing literature (4, 9–13) as well as our own fluoroscopic images of 8 whole-breast radiotherapy patients with clips, 5 of whom were also among our CBCT study patients. The postcorrection PTV margin was calculated as qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi P 2:33 ð ðmi Þ2 þRMSðsi Þ2 Þ. Because the population standard deviaqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi P P tion, p ; z ð ðmi Þ2 þRMSðsi Þ2 Þ, this margin is roughly equivalent to that obtained using the two-parameter model at a 98% confidence level (8).
RESULTS Individual patient mean setup errors (m) and standard deviations (s) are shown in Fig. 1. The precorrection mean and standard deviation of the individual mean setup errors, M(mi) S(mi), were 2.7 4.3, 1.4 3.1, and 2.5 2.5 mm in the right–left (RL), anterior–posterior (AP), and superior–inferior (SI) directions, respectively. The root mean square and standard deviation of the individual standard deviations, RMS(si) S(si), before correction were 2.2 1.4, 1.7 0.4, and 1.6 0.7 mm in the RL, AP, and SI directions, respectively. After correction, the mean and standard deviation of the individual mean residual errors, M(mi) S(mi), were 0.2 0.9, 0.4 1.4, and 0.0 1.1 mm in the RL, AP, and SI directions, respectively. The root mean square and standard deviation of the individual standard deviations, RMS(si) S(si), after correction were 0.6 0.2, 1.0 0.5, and 0.7 0.3 mm in the RL, AP, and SI directions, respectively. The global mean and standard deviation of surrogate errors were calculated to be 0.4 2.4, 0.6 1.7, and 0.3 2.3 mm for RL, AP, and SI directions, respectively, for the 12 patients from the Weed et al. study with imaging dates consistent with the NSABP B-39/RTOG 0413 protocol. The CTV-to-PTV margin allowable when clip-based, online setup correction is used was calculated to be 6 mm in each orthogonal direction.
Lumpectomy cavity localization using clips d L. H. KIM et al.
Fig. 1. Patient-specific setup errors, before and after correction (mean and 1 standard deviation). Ant = anterior; Post = posterior; Sup = superior; Inf = inferior.
DISCUSSION Setup correction procedure The study patients who underwent this procedure received whole-breast irradiation from two tangent beams. In that situation it was easy to set up the patient and couch to allow both treatment and a 200 CBCT scan without any collision issue. In more recent experience with actual image-guided APBI treatments, we found it strongly advisable, before the beginning of treatment, to assess whether such a setup is possible (i.e., a setup whereby all scans, portal images, and treatment geometries can be executed without having to move the couch or patient because of clearance issues). We found that the lateral and longitudinal couch positions may need to be narrowly specified to achieve this. An alternative solution is to scan the patient in an out-of-treatment position that physically allows the CBCT scan while ensuring that the target and other relevant structures remain in the field of view for registration purposes. Registration is performed to determine the ‘‘true’’ treatment position and the necessary couch ‘‘correction’’ required to reach it. The CBCT components are then retracted (withdrawing the couch if necessary to do this) before the couch is moved to the calculated treatment position. We have found this method to work well, with one drawback being that a postcorrection verification CBCT cannot be acquired in actual treatment position. PTV margin calculation Our precorrection setup data suggest that, without setup correction, margins of 12, 9, and 9 mm are required for 98% confidence in the RL, AP, and SI directions, respectively, if M(mi) = 0. Although we obtained values for M(mi)
1307
of 2.7, 1.4, and 2.5 mm in the RL, AP, and SI directions, respectively, our data are not inconsistent with M(mi) = 0 because of the magnitude of S(si) and our small patient sample, with the possible exception of SI. If M(mi) is really non-zero, this indicates a miscalibration between simulation and treatment setups that should be corrected before application of the quoted margins. Assuming M(mi) = 0, the 10-mm CTVto-PTV margin called for by the NSABP B-39/RTOG 0413 protocol (7) seems appropriate, with the possible exception of LR, for which the observed errors suggest that a 10-mm margin offers only 94% confidence. For this direction a margin of 12 mm or more might be considered to attain 98% confidence. This analysis assumes that there is negligible difference in precorrection setup error statistics of whole-breast patients, who made up our study population, and APBI patients. At our institution, at least, this seems reasonable because the patients are tattooed and set up identically. Further support for this assumption comes from Gierga et al. (2), who measured setup error after laser-based setup of APBI patients and obtained errors of the same magnitude as ours. According to our calculations, the use of our setup correction procedure would allow a CTV-to-PTV margin of 6 mm, a 4-mm reduction of the margin specified in NSABP B-39/ RTOG 0413. As noted above, the protocol value may even be low in one direction. Baglan et al. (10) obtained the 10mm figure partly on the basis of on an anticipated setup error of 5 mm. Our results and those of Gierga et al. suggest that this may have been slightly optimistic. On the other hand, Baglan et al. allowed 5 mm of margin for breathing motion, which is more than what we allow. It must be noted that estimates of breathing motion for APBI patients vary widely in the literature, ranging from 1 to 10 mm full excursion (4, 9– 13). Baglan et al.’s measurements are among the highest in the literature. Examination of their data suggests that their method of manually registering three CT scans to obtain range-of-motion estimates could have contributed to the larger values. Nevertheless, there is disagreement in the literature regarding magnitude of breathing motion, and its resolution could affect our margin calculation. Regarding margin calculation, we noted above that the patient data from the Weed et al. study do not allow determination of whether surrogate error is systematic or random. When one considers possible causes of surrogate error—contouring, clip migration, and sampling (i.e., clips are a small, nonrandom sample of the cavity)—it seems plausible that surrogate error for an individual patient is largely systematic. Ongoing work in which 6 APBI patients to date have received multiple CT scans of clips and cavity indicates that this is probably true. In that case, it may be possible to account for surrogate error largely through correction rather than the PTV margin. For example, a CT scan acquired just before the start of a patient’s treatment course could be used with the planning CT to estimate the magnitude and direction of that patient’s surrogate error. That can be addressed with an additional, global couch correction for every treatment fraction. The PTV margin would then need only include random variation of
I. J. Radiation Oncology d Biology d Physics
1308
Volume 69, Number 4, 2007
Table 1. Comparison with results from White et al.* Precorrection
Postcorrection
Parameter
LR
AP
SI
LR
AP
SI
White systematic Kim S(mi)
2.7 4.3
1.7 3.1
2.4 2.5
0.8 0.9
0.7 1.4
0.8 1.1
2.4 2.2 (1.4)
2.2 1.7 (0.4)
2.9 1.6 (0.7)
1.5 0.6 (0.2)
1.5 1.0 (0.5)
1.6 0.7 (0.3)
8.8 11.3
8.8 8.2
8.8 6.8
3.6 2.4
3.6 4.0
3.6 3.1
White random Kim RMS(si) (s(si)) White margin Kim margin (setup only)
Abbreviations: LR = left–right; AP = anterior–posterior; SI = superior–inferior. Values are millimeters. * Reference 14.
surrogate error—for the 6 APBI patients mentioned above, the RMS(si) of surrogate errors was approximately 0.8 mm. It should be emphasized that the just-described procedure is at present speculative and untested. A recent study by White et al. (14) describes CBCT-based, on-line APBI setup correction using manual registration of the chest wall and skin. A comparison of their results with ours is given in Table 1. For this comparison, surrogate error and breathing were removed from our margin estimates because these were not included in White et al. The two studies seem generally consistent despite the use of different surrogate structures and localization methods. This suggests that global setup error is presently a relatively greater issue than internal target motion for treatment accuracy. This hypothesis is supported by the observation that couch corrections based on clips vs. those based on whole-breast registration are well correlated (Fig. 2; r = 0.8) with a mean difference between a clip-based correction and a whole-breast-based correction in a given direction of 2.2 1.5 mm. These data
Fig. 2. Correlation between clip-based and whole-breast-based couch corrections.
are based on 5 of our study patients (24 images) for whom both types of registration were performed. The agreement in post-correction measurements in the two studies supports our assumption that the use of whole-breast patients rather than APBI patients does not affect the measurement of residual error. One of the key assumptions in this study is that the cavity COM is a suitable surrogate for the CTV. The CTV is normally generated by applying a margin to the lumpectomy cavity as visualized in the planning CT. Numerous studies have shown that the lumpectomy cavity’s shape and volume can change significantly over time (4, 15–20). Even within the comparatively short time frames of APBI, large changes have been observed. For the 12 patients in the Weed et al. study with imaging dates consistent with the RTOG 0413/ NSABP B-39 protocol, the mean volume reduction was 45% between planning and localization CTs. In recent work at our institution we observed a volume reduction of 30% 35% between planning and treatment in 16 APBI patients. In addition, volume changes were observed to be generally non-isotropic.
Fig. 3. Example of nonuniform cavity shrinkage.
Lumpectomy cavity localization using clips d L. H. KIM et al.
It is not clear how and to what extent the CTVs change with the cavity. If the CTV is essentially static, regardless of cavity volume change, using surgical clips for setup correction would be inadvisable because surgical clips track the cavity. On the other hand, if CTVs do change with the cavity over time, the nonuniformity of this change still raises issues given that the cavity COM is moving because of local volume changes, not target translation. For example, Fig. 3 shows cavity contours from two CTs acquired 3 weeks apart. The shrinkage is clearly non-uniform— although the smaller contour is bounded by the original, the center of the cavity has moved 12 mm between the two CTs. Yet it may be preferable to leave the setup as is and not make a cavity COMbased couch correction of 12 mm in the direction of the lung. If volume changes result in large relative shifts between the cavity and other structures, then global localization using
1309
the whole breast, surface, or chest wall might be preferable to cavity localization.
CONCLUSIONS On-line localization of the lumpectomy cavity using surgical clips is technically feasible from the standpoint of equipment, time, and process. Using this procedure, PTV margins for external beam APBI can be reduced to approximately 6 mm. Further reduction would require better performance and characterization of the target surrogate. Because this image guidance procedure is exclusively target based, additional monitoring of organs at risk may be advisable. Alternatively, localization based on the whole breast and chest wall could be considered.
REFERENCES 1. Langen K, Buchholz D, Victor G, et al. Evaluation of breast tumor bed alignment techniques for image guided radiotherapy [abstract]. Int J Radiat Oncol Biol Phys 2005;63(Suppl. 1): S179. 2. Gierga DP, Turcotte JC, Riboldi M, et al. Comparison of target registration errors for multiple modalities in image-guided partial breast irradiation [abstract]. Int J Radiat Oncol Biol Phys 2006;66(Suppl. 1):S103–S104. 3. Hasan Y, Kim L, Weed D, et al. Image guidance in external beam accelerated partial breast irradiation: Comparison of surrogates for the lumpectomy cavity [abstract]. Int J Radiat Oncol Biol Phys 2005;63(Suppl. 1):S138. 4. Weed DW, Yan D, Martinez AA, et al. The validity of surgical clips as a radiographic surrogate for the lumpectomy cavity in image-guided accelerated partial breast irradiation. Int J Radiat Oncol Biol Phys 2004;60:484–492. 5. Wong J, Watt L, Yan D, et al. ARTIST: Work flow for image guided radiation therapy. In: Yi BY, Ahn SD, Choi EK, et al., editors. Proceedings of the XIVth International Conference on the Use of Computers in Radiation Therapy; 2004 May 1013; Seoul, South Korea. Seoul: Jeong; 2004. p. 417–420. 6. Kim L, Cernej R, Vargas C, et al. Reduction of PTV margin for accelerated partial breast irradiation using on-line detection of surgical clips [virtual poster]. 2004 ASTRO Annual Meeting Abstracts and Virtual Posters [web site]. Fairfax, VA: American Society for Therapeutic Radiology and Oncology; 2004. Available at: http://astro.abstractsnet.com/handouts/000132_1133post. pdf. Accessed June 5, 2006. 7. National Surgical Adjuvant Breast and Bowel Project, Radiation Therapy Oncology Group. A randomized phase III study of conventional whole breast irradiation (WBI) versus partial breast irradiation (PBI) for women with stage 0, I, or II breast cancer. Philadelphia: Radiation Therapy Oncology Group; 2006. Available from: http://www.rtog.org/members/protocols/ 0413/0413.pdf. Accessed October 24, 2006. 8. Yan D, Lockman D, Martinez A, et al. Computed tomography guided management of interfractional patient variation. Semin Radiat Oncol 2005;15:168–179. 9. Bert C, Metheany KG, Doppke KP, et al. Clinical experience with a 3D surface patient setup system for alignment of partial-breast irradiation patients. Int J Radiat Oncol Biol Phys 2006;64:1265–1274. 10. Baglan KL, Sharpe MB, Jaffray D, et al. Accelerated partial breast irradiation using 3D conformal radiation therapy (3DCRT). Int J Radiat Oncol Biol Phys 2003;55:302–311.
11. Lee R, Suh H, Moon B, et al. Evaluation of breast motion as assessed by breast clips during normal breathing. In: Yi BY, Ahn SD, Choi EK, et al., editors. Proceedings of the XIVth International Conference on the Use of Computers in Radiation Therapy; 2004 May 10-13; Seoul, South Korea. Seoul: Jeong; 2004. p. 442–444. 12. Chopra S, Dinshaw KA, Kamble R, et al. Breast movement during normal and deep breathing, respiratory training and set up errors: Implications for external beam partial breast irradiation. Br J Radiol 2006;79:766–773. 13. Yue NJ, Li X, Beriwal S, et al. The intrafraction motion induced dosimetric impacts in breast 3D radiation treatment: A 4DCT based study. Med Phys 2007;34:2789–2800. 14. White EA, Cho J, Vallis KA, et al. Cone beam computed tomography guidance for setup of patients receiving accelerated partial breast irradiation. Int J Radiat Oncol Biol Phys 2007; 68:547–554. 15. Goldberg H, Prosnitz R, Olson J, et al. Definition of postlumpectomy tumor bed for radiotherapy boost field planning: CT versus surgical clips. Int J Radiat Oncol Biol Phys 2005;63: 209–213. 16. Oh KS, Kong F, Griffith KA, et al. Planning the breast tumor bed boost: Changes in the excision cavity volume and surgical scar location after breast-conserving surgery and whole-breast irradiation. Int J Radiat Oncol Biol Phys 2006;66:680–686. 17. Jacobson G, Betts V, Smith B. Change in volume of lumpectomy cavity during external-beam irradiation of the intact breast. Int J Radiat Oncol Biol Phys 2006;65:1161–1164. 18. Kubicek G, Kimler B, Bell M, et al. Changes in breast cavity contours after lumpectomy [virtual poster]. 2006 ASTRO Annual Meeting Abstracts and Virtual Posters [web site]. Fairfax, VA: American Society for Therapeutic Radiology and Oncology; 2006. Available from: http://astro2006.abstractsnet.com/ aoriginal.wcs?entryid=000060. Accessed February 20, 2007. 19. Vicini F, Remouchamps V, Wallace M, et al. Ongoing clinical experience utilizing 3D conformal external beam radiotherapy to deliver partial-breast irradiation in patients with early-stage breast cancer treated with breast-conserving therapy. Int J Radiat Oncol Biol Phys 2003;57:1247–1253. 20. Kader H, Truong P, Panades M, et al. Is breast seroma the appropriate guide for PTV determination in accelerated partial breast irradiation planning? [Abstract]. Radiother Oncol 2004; 72(Suppl.):S6.