Geometrical variability of esophageal tumors and its implications for accurate radiation therapy - Chapter 3: Marker-based quantification of interfractional tumor position variation and the use of markers for setup verification

Hele tekst

(1)UvA-DARE (Digital Academic Repository). Geometrical variability of esophageal tumors and its implications for accurate radiation therapy Jin, P. Publication date 2019 Document Version Other version License Other Link to publication Citation for published version (APA): Jin, P. (2019). Geometrical variability of esophageal tumors and its implications for accurate radiation therapy.. General rights It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons). Disclaimer/Complaints regulations If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible.. UvA-DARE is a service provided by the library of the University of Amsterdam (https://dare.uva.nl) Download date:21 Jun 2021.

(2) 3. Marker-based quantification of interfractional tumor position variation and the use of markers for setup verification in radiation therapy for esophageal cancer. P. Jin, A. van der Horst, R. de Jong, J.E. van Hooft, M. Kamphuis, N. van Wieringen, M. Machiels, A. Bel, M.C.C.M. Hulshof, and T. Alderliesten. A version of this chapter has been published in Radiotherapy and Oncology. 2015; 117(3): 412–418 DOI: 10.1016/j.radonc.2015.10.005. 527155-L-bw-Jin Processed on: 18-12-2018. PDF page: 39.

(3) Chapter 3. Abstract Purpose The aim of this study was to quantify interfractional esophageal tumor position variation using markers and investigate the use of markers for setup verification.. Materials and methods Sixty-five markers placed in the tumor volumes of 24 esophageal cancer patients were identified in computed tomography (CT) and follow-up cone-beam CT. For each patient we calculated pairwise distances between markers over time to evaluate geometrical tumor volume variation. We then quantified marker displacements relative to bony anatomy and estimated the variation of systematic (Σ) and random errors (σ). During bony anatomy-based setup verification, we visually inspected whether the markers were inside the planning target volume (PTV) and attempted marker-based registration.. Results Minor time trends with substantial fluctuations in pairwise distances implied tissue deformation. Overall, Σ(σ) in the left–right/cranial–caudal/anterior–posterior direction was 2.9(2.4)/4.1(2.4) /2.2(1.8) mm; for the cardia, it was 5.4(4.3)/4.9(3.2)/1.9(2.4) mm. After bony anatomy-based setup correction, all markers were inside the PTV. However, due to large tissue deformation, marker-based registration was not feasible.. Conclusions Generally, the interfractional position variation of esophageal tumors is more pronounced in the cranial–caudal direction and in the cardia. Currently, marker-based setup verification is not feasible for clinical routine use, but markers can facilitate the setup verification by inspecting whether the PTV covers the tumor volume adequately.. 40. 527155-L-bw-Jin Processed on: 18-12-2018. PDF page: 40.

(4) Interfractional position variation in esophageal cancer RT. 3.1 Introduction With the most rapidly increasing incidence [155, 156], esophageal cancer has been estimated globally as the eighth most common cancer and the sixth most common cause of death from cancer [1]. Neoadjuvant and definitive chemoradiation therapy are the preferred treatment modalities for resectable and unresectable/inoperable (gastro-)esophageal cancer patients [17, 18, 157]. Currently, in clinical image-guided radiation therapy (IGRT) for esophageal cancer, it is common to rigidly register the three-dimensional (3D) planning computed tomography (pCT) or two-dimensional (2D) digitally reconstructed radiographs (DRRs) with the kilo-/megavoltage (kV/MV) cone-beam CT (CBCT) or 2D fluoroscopy images on bony anatomy (i.e., the vertebrae) for patient setup verification [63, 115, 116, 158, 159]. Although the actual tumor volumebased registration is preferred, it is virtually impossible due to the limited soft-tissue contrast in CT and CBCT. Hence, delineation uncertainties and intra-/interfractional tumor position variation relative to bony anatomy currently prompt the use of large isotropic safety margins for uncertainty compensation [17]. However, this can lead to potential toxicities in organs at risk [17, 160] and hamper the use of dose-escalation for improving locoregional control of definitive chemoradiation therapy [20].. 3. For a number of tumor sites, fiducial markers have successfully aided delineation, tumor position variation quantification, and tumor-based setup verification [161–164]. For esophageal tumors, endoscopy-/endoscopic ultrasound (EUS)-guided marker placement was also found feasible and useful for accurately projecting the gross tumor volume (GTV) extent onto the planning CT [61–63]. However, few studies have quantified the intra-/interfractional position variation of esophageal tumors using fiducial markers [159, 165]. Moreover, the potential benefit of using markers for patient setup verification in IGRT for esophageal cancer has not yet been investigated. In this study, we included esophageal cancer patients with markers placed in the tumor volume and manually identified these markers in the pCT and follow-up CBCT scans. We aimed to quantify the interfractional position variation of esophageal tumors relative to bony anatomy using the markers. In addition, we investigated the use of markers for patient setup verification.. 3.2 Materials and methods Patient and marker characteristics From March 2013 to May 2014, we consecutively included 30 esophageal cancer patients (24 males and 6 females) aged 45–84 (average: 66) years in our study. This patient population is identical to the one in a pilot study concerning the feasibility of marker placement [63]. For one patient, markers failed to be placed due to a manufacturing error in the preloaded needle system; all other 29 patients underwent successful endoscopy-/EUS-guided marker placement prior to. 41. 527155-L-bw-Jin Processed on: 18-12-2018. PDF page: 41.

(5) Chapter 3 Table 3.1: Overview of patient and marker characteristics.. Patient. Tumor Tumor type location. Marker type. Marker length/ volume. Dose scheme [Gy]. No. of CBCTs. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30. AC AC AC PDC SCC SCC AC AC SCC AC SCC SCC SCC SCC SCC AC AC AC AC AC AC AC AC AC AC SCC SCC SCC SCC AC. Solid Solid Solid Solid Solid Flexible Flexible Flexible Solid Flexible Flexible Flexible Hydrogel Hydrogel Hydrogel Flexible Hydrogel Flexible Flexible Flexible Flexible Flexible Flexible Flexible Hydrogel Solid Solid Solid Solid. 5 mm 5 mm 5 mm 5 mm 5 mm 5–10 mm 3 mm 4 mm 5 mm 2–10 mm 10 mm 7–8 mm 0.40 ml 0.40 ml 0.40 ml 10 mm 0.40 ml 8 mm 8 mm 8 mm 8 mm 10 mm 10 mm 10 mm 0.40 ml 5 mm 5 mm 5 mm 5 mm. 23 × 1.8 23 × 1.8 23 × 1.8 23 × 1.8 28 × 2.2 28 × 1.8 23 × 1.8 28 × 1.8 23 × 1.8 23 × 1.8 23 × 1.8 28 × 1.8 23 × 1.8 28 × 1.8 23 × 1.8 28 × 1.8 23 × 1.8 23 × 1.8 23 × 1.8 23 × 1.8 28 × 1.8 23 × 1.8 23 × 1.8 23 × 1.8 23 × 1.8 23 × 1.8 28 × 1.8 28 × 1.8 23 × 1.8 23 × 1.8. Distal Distal Distal Distal Distal Distal Distal Distal Middle Distal Distal Proximal Middle Proximal Distal Distal Distal Distal Distal Distal Distal Distal Distal Distal Distal Distal Proximal Middle Distal Full. Total. No. of markers At placement. Visible in pCT. Visible in CBCTs. 7 7 7 7 8 8 8 28 8 7 12 8 23 25 8 8 8 23 7 11 8 9 12 12 8 8 8 9 9 7. 2‡ 3‡ 3‡ 3 0 4 4 5 4 3 3 3 4 3 6 3 3 3 3 3 4 4 3 4 4 3 3 5 3 3. 1 2 3 3 0 4 4 5 3 2 2† 3 4 3 5 3 3 3 3 2† 4 4 3 4 4 2 3 4† 2 3. 1 1 1 3→2 0* 4 4 0* 3 0* 2† 3 4 1 0* 3→1 3 0 3→2 2† 3 4 3 2 4 0* 3 4 2 2. 318. 101. 91. 65 → 61. Diameter of gold markers: solid: 0.43–0.64 mm or 0.35–0.50 mm; flexible: 0.35 mm. Arrow (→) means marker went missing during the treatment course. Abbreviations: AC = adenocarcinoma, SCC = squamous cell carcinoma, PDC = poorly differentiated carcinoma; pCT = planning computed tomography, CBCT = cone-beam computed tomography. * No marker was visible in CBCT; therefore these patients were excluded from data analysis. † Compared to the pilot study [63], there is a difference of 1 in the number count because we excluded a metal clip (patient 11), a marker located in the lung in the pCT and CBCT (patient 20), or a marker that detached between implantation and acquisition of the pCT and was therefore located in the stomach in the pCT (patient 28). ‡ Two markers placed in the same location by accident have been counted as one marker [63].. pCT acquisition. The medical ethics committee of our institute approved the marker implantation and all patients gave written informed consent [63]. Table 3.1 lists the patient and marker characteristics. Two different types of gold markers and one gel-based marker were used: solid marker (Cook Medical, Limerick, Ireland; or in-house manufactured), flexible coil-shaped marker (Visicoil; IBA Dosimetry, Bartlett, TN), and hydrogel marker (TraceIt; Augmenix, Waltham, MA). For each patient, we placed at least 2 markers of the same type, preferably in the submucosal layer at the cranial and caudal border and in the center of the primary tumor, as described in [63]. For 5 patients, no markers were identified in any of the CBCT scans, due to marker detachment after the placement, too short hand-cut flexible coil-shaped marker, or absorption/dissolution of hydrogel in the tissue [63]. Therefore, 24 patients with in total 65 markers with clear visibility in CBCT were included in our data analysis (Table 3.1). The 65 markers were classified, according to the American Joint Committee on Cancer manual [15], into four subgroups based on their locations. 42. 527155-L-bw-Jin Processed on: 18-12-2018. PDF page: 42.

(6) Interfractional position variation in esophageal cancer RT. in the esophagus: 12, 11, 31, and 11 markers, in the proximal esophagus, middle esophagus, distal esophagus, and cardia, respectively (Fig. 3.1).. n=12. Proximal esophagus. Carina. n=11. Middle esophagus. Inferior pulmonary vein. 3 n=31. Distal esophagus. Gastroesophageal junction n=11. Cardia. Fig. 3.1: Schematic representation of positions of all 65 markers (filled dots) placed in the esophagus (coronal view).. Image acquisition and target delineation For each patient, a 3D pCT scan was acquired within five days (average: one day) after marker placement. During pCT acquisition (LightSpeed RT 16 CT; General Electric Company, Waukesha, WI), all patients were positioned supine with arms up above their heads using an arm support (CIVCO Medical Solution, Coralville, IA). The thickness of the axial scan slices was 2.5 mm, and the field of view was between the bottom edge of the mandible and the lower border of the kidneys. The GTV was contoured on the pCT by the radiation oncologist with the aid of markers and all diagnostic information including physical examination, diagnostic CT, endoscopy, and EUS. The clinical target volume (CTV) was the extension of the GTV with a 35-mm cranial–caudal (CC) margin and a 5-mm circumferential margin. Anatomical boundaries and periesophageal lymph nodes were also taken into account. The CTV to planning target volume (PTV) margin was 10 mm in all directions. Following the extended no action level protocol (eNAL) [114], a total of 7–8 CBCT scans per patient were acquired (Elekta Synergy System; Elekta AB, Stockholm, Sweden) for setup verification: daily CBCT acquisition for the first 4 fractions, followed by once-weekly acquisitions. For the fractions without CBCT, patients were positioned based on the average setup error calculated using the available CBCTs. More CBCT scans were acquired when the results of eNAL exceeded. 43. 527155-L-bw-Jin Processed on: 18-12-2018. PDF page: 43.

(7) Chapter 3. tolerance (e.g., patients 13 and 14). The 236 CBCT scans of the 24 patients with visible markers were included in our analysis (Table 3.1).. Pairwise distance between markers over treatment course For the 19 patients with more than one marker visible in the last CBCT (in total 61 marker pairs), we calculated the variation of pairwise distances between markers to investigate geometrical variations of the tumor volume over time. Using X-ray Volume Imaging (XVI) software (Elekta), we retrieved for each patient the coordinates of the marker centers in the pCT and each CBCT and derived the pairwise Euclidian distances. Further, we applied linear regression analysis on the pairwise distances over the treatment course for a coarse examination of time trend. The slopes and residuals of the linear fits were then analyzed.. Interfractional marker displacement To quantify the position variation of esophageal tumors, we calculated the individual marker displacement relative to bony anatomy in the left–right (LR), CC, and anterior–posterior (AP) direction, and also in 3D vector distance. As they may be dependent on location due to the elongated shape of the esophagus, marker displacements were analyzed for the whole marker group and the four marker subgroups: markers located in the proximal esophagus, middle esophagus, distal esophagus, and cardia. Retrospectively, we rigidly registered for each patient each CBCT to the pCT based on the vertebrae using XVI. We then calculated for each CBCT the displacement of each marker relative to its corresponding marker position in the pCT. The mean and standard deviation (SD) of interfractional marker displacements were subsequently calculated, which are estimates of the systematic error and the SD of the random errors for individual markers, respectively. Further, for the whole marker group and the four marker subgroups we estimated the group mean (M, the mean of systematic errors), the SD of systematic errors (Σ), and the root mean square of SDs of random errors (σ) [134]. Subsequently, we calculated the margins required to compensate for only the interfractional position variation of esophageal tumors based on the margin recipe 2.5Σ + 0.7σ [134].. Statistical analysis To compare the interfractional displacements between the three orthogonal directions and between the four marker subgroups, we applied a Friedman test with Wilcoxon signed-rank test and a Kruskal-Wallis test with Dunn’s test, respectively. Holm adjustment was performed to all post hoc tests. Results with p<0.05 were considered to be significant. All statistical analyses were performed using the R software package (version 3.0.2) [144].. 44. 527155-L-bw-Jin Processed on: 18-12-2018. PDF page: 44.

(8) Interfractional position variation in esophageal cancer RT. 3. Fig. 3.2: Illustration of the automatic marker-based registration procedure (patient 19). On the sagittal scan provided here, markers on the planning computed tomography (CT, purple) and cone-beam CT (CBCT, green) are indicated by the purple and green arrows, respectively. (a) Overlay of the original planning CT and CBCT (tattoo-based alignment); (b) overlay after bony anatomy registration; (c and d) overlay after subsequent automatic marker-based registration, with (c) and without (d) visualization of the mask used for marker-based registration.. Setup verification with markers Retrospectively, each CBCT was rigidly registered to the pCT based on the vertebrae by use of XVI, as is routinely done for setup verification in our clinic. After correcting the setup position, it was subsequently examined for all identified markers whether they were inside the GTV/CTV/ PTV delineations. To test whether a set of markers can be used as a surrogate for setup verification, we then assessed the feasibility of marker-based registration using XVI. More specifically, after bony anatomy registration, a 3D GTV-shaped mask (i.e., a region of interest) covering all markers in the pCT was generated, followed by an attempt of automatic rigid image registration using the Chamfer-matching or the gray-value-matching technique (example: Fig. 3.2) [166]. When the automatic registration failed, a manual rigid registration was attempted by altering the translations and, if necessary, the rotations. All the registrations in our study were done by an experienced radiation therapist (R.d.J.).. 3.3 Results The range of pairwise distances between markers was large (in pCT: 9.5–177.9 mm, n = 61). The mean ± SD of slopes of linear fits was −0.04 ± 0.16 mm/day, implying minor time trends. Moreover, 36 (59%) marker pairs showed no significant difference from 0 mm/day in slope. The me-. 45. 527155-L-bw-Jin Processed on: 18-12-2018. PDF page: 45.

(9) Chapter 3. 140 80. SD[mm] M1-2: 1.04 M1-3: 1.05 M2-3: 0.76 M1-4: 0.74 M2-4: 0.96 M3-4: 1.17. 40 20. 20. 40. 40. 60. 80 60. Slope[mm/day] M1-2: 0.19** M1-3: 0.18** M2-3: 0.00 M1-4: 0.28*** M2-4: 0.13* M3-4: 0.15*. 120. SD[mm] M1-2: 1.51 M1-3: 3.81 M2-3: 2.71 M1-4: 1.78 M2-4: 1.47 M3-4: 3.44. 100. 100. Slope[mm/day] M1-2: 0.07* M1-3: 0.27** M2-3: -0.01 M1-4: 0.07 M2-4: -0.05 M3-4: -0.13. 80. SD[mm] M1-2: 1.56 M1-3: 2.23 M2-3: 1.52 M1-4: 2.82 M2-4: 2.43 M3-4: 3.11. 60. 100. 120. Slope[mm/day] M1-2: -0.17* M1-3: -0.41** M2-3: -0.23* M1-4: -0.35* M2-4: -0.19 M3-4: -0.34*. 20. Pairwise distance between markers [mm]. 140. dian and range of the SDs of residuals were 1.5 mm and 0.4–6.7 mm, respectively. In addition, visual inspection on the pairwise distances demonstrated substantial fluctuations over the treatment course (example: Fig. 3.3).. 10. 20. 30. 40. Number of days since pCT. 50. 0. Patient 25, four markers. 0. Patient 13, four markers. 0. Patient 7, four markers 0. 0. 10. 20. 30. 40. Number of days since pCT. 50. 0. 10. 20. 30. 40. 50. Number of days since pCT. Fig. 3.3: The pairwise distances between markers over the treatment course (with linear fitted lines) for patients 7, 13, and 25. Slope: slope of fit, with significance code (* p<0.05, ** p<0.01, *** p<0.001); SD: standard deviation of residuals. Note: vertical axes differ in scale; M1–2 denotes the pair consisting of markers 1 and 2.. Figure 3.4 shows for each of the 65 markers the interfractional displacements relative to bony anatomy in the LR, CC, and AP direction as well as in 3D vector distance. We found that 12% and 49% of all 613 3D vector marker displacements were more than 10 mm and 5 mm, respectively. Fig. 3.5 illustrates for each marker the systematic error and the SD of the random errors. For the four marker subgroups and the whole marker group, the M, Σ, σ, and the estimated margin in the LR, CC, and AP direction are listed in Table 3.2. For the group of all 65 markers and subgroup of markers located in the distal esophagus (n = 31), the absolute systematic errors and SDs of random errors were significantly larger in the CC direction than in the LR and AP direction (p<0.05). No such significant difference was found for the subgroups of markers located in the proximal esophagus, middle esophagus, and cardia (Fig. 3.6). Further, significantly larger systematic errors and SDs of random errors in 3D vector distance were obtained for the markers located in the cardia than in the proximal, middle, and distal esophagus (p<0.05). However, we found no significant differences between the proximal, middle, and distal esophagus (Fig. 3.6). After bony anatomy-based setup correction, the identified markers remained inside the GTV, CTV, and PTV delineations in 420 (69%), 582 (95%), and 613 (100%) cases of inspections, respectively. Automatic marker-based registration was only feasible for in total 21 CBCT scans: 2 of patient 4, 8 of patient 6, 7 of patient 19, and 4 of patient 27. For one CBCT scan of patient 13, neither the automatic nor the manual marker-based registration succeeded, because the marker. 46. 527155-L-bw-Jin Processed on: 18-12-2018. PDF page: 46.

(10) 0 -20 -10. Cranial–Caudal. 20 10 0. Anterior–Posterior. -20 -10 20 10 0 -20 -10. Proximal esophagus Middle esophagus Distal esophagus Cardia 1 2 3. 4. 6. 3. 3D vector. 20 10 0 -20 -10. Marker displacement relative to bony anatomy [mm]. 10. Left–Right. 20. Interfractional position variation in esophageal cancer RT. Systematic (mean) displacement 7. 9. 11. 12. 13. 14. 16. 17. 19. 20. 21. 22. 23. 24. 25. 27. 28. 29. 30. Patient number. Fig. 3.4: For each marker (n = 65; 24 patients), the interfractional displacements relative to bony anatomy in the left– right, cranial–caudal, and anterior–posterior directions as well as the three-dimensional (3D) vector distance (from the top to the bottom panel).. 1 mm Proximal esophagus. n=12. n=11. Middle esophagus. n=31. Distal esophagus. Cranial. Cranial Cardia. Right. Posterior. n=11. Left. Anterior Caudal. Caudal Coronal View. Sagittal View. Fig. 3.5: Illustration of the systematic error (the length of the arrow) and the standard deviation (SD) of random errors (the length of semi-major/minor axis of the ellipse) for each marker, projected onto the coronal (left) and sagittal (right) views of the esophagus. Note: the amplitudes of the errors are not scaled to the esophagus.. 47. 527155-L-bw-Jin Processed on: 18-12-2018. PDF page: 47.

(11) Chapter 3. Table 3.2: The group mean (M), standard deviation (SD) of systematic errors (Σ), root mean square of SDs of random errors (σ), and the estimated margin required to compensate for the interfractional tumor position variation in the. CC [mm]. AP [mm]. M Σ σ 2.5Σ + 0.7σ. 0.1 1.5 1.3 4.4. −1.2 4.1 1.5 11.3. −0.7 1.9 1.2 5.5. Middle esophagus (n = 11). M Σ σ 2.5Σ + 0.7σ. −1.1 3.1 1.5 8.9. −2.3 2.9 2.0 8.6. 1.2 3.2 2.3 9.6. Distal esophagus (n = 31). M Σ σ 2.5Σ + 0.7σ. 0.4 1.9 1.9 6.1. −0.8 4.2 2.5 12.1. 0.5 1.9 1.4 5.7. Cardia (n = 11). M Σ σ 2.5Σ + 0.7σ. −1.1 5.4 4.3 16.4. −1.2 4.9 3.2 14.6. −0.1 1.9 2.4 6.4. Overall (n = 65). M Σ σ 2.5Σ + 0.7σ. −0.2 2.9 2.4 -. −1.2 4.1 2.4 -. 0.3 2.2 1.8 -. 15. LR [mm] Proximal esophagus (n = 12). **. *. ***. 5 0. Upper and lower quartiles Highest/lowest data within 1.5 × interquartile range of the upper/lower quartiles. Median LR,CC,AP Median 3D Outlier Significant difference. *. ** ***. *. *. ***. 5. 10. **. *. **. Boxes: Whiskers:. 0. 3D. AP. LR. CC. Overall (n=65) 3D. AP. LR. CC. Cardia (n=11) 3D. AP. LR. CC. Distal esophagus (n=31) 3D. AP. LR. CC. 3D. AP. LR. Middle esophagus (n=11). *. **. ***. Proximal esophagus (n=12) CC. SD of random errors [mm]. **. **. 10. **. 15. Absolute systematic error [mm]. left–right (LR), cranial–caudal (CC), and anterior–posterior (AP) direction for the four subgroups of markers.. Direction. Fig. 3.6: Distributions of absolute systematic errors and standard deviations of random errors of individual markers, for each marker subgroup separately as well as for all markers together (Overall). Results are given in the left–right (LR), cranial–caudal (CC), and anterior–posterior (AP) direction as well as in the three-dimensional vector distance (3D). Significance code: * p<0.05, ** p<0.01, *** p<0.001.. 48. 527155-L-bw-Jin Processed on: 18-12-2018. PDF page: 48.

(12) Interfractional position variation in esophageal cancer RT. positions relative to each other were greatly different between the CBCT and the pCT. For the remaining 214 CBCT scans, we attempted a manual marker-based registration, but it was very time-consuming, extremely challenging, and difficult to evaluate the result since the tumor volume deformed considerably.. 3.4 Discussion In this retrospective study, we investigated the interfractional position variation of esophageal tumors relative to bony anatomy assisted by markers and the potential use of markers in the setup verification process. Large systematic and random errors were predominantly found in the CC direction and in the cardia. Further, we found that markers can facilitate setup verification by inspecting whether the PTV covers the tumor volume adequately. For the pairwise distances between markers over the treatment course, the linear fit could at least give a rough estimation of a possible time trend, even though the normality and homoscedasticity of the residuals could not be powerfully tested due to the small sample size. The pairwise distance can be affected by marker registration error, marker migration, and tissue deformation. Marker registration errors for the flexible coil-shaped markers in CT and CBCT using XVI were found to be less than 0.8 mm [161], which is small compared to the SDs of the residuals we found. Inherent to the method of pair distance analysis, marker migration and tissue deformation cannot be distinguished. However, since the gold markers were placed superficially, potential marker migration would likely result in marker detachment [63]. In addition, migration would not likely be characterized by alternating between increase and decrease of pairwise distance. Therefore, tissue deformation most likely explains the observed substantial fluctuation in the pairwise distances of markers over the treatment course, also considering the long extent of the primary tumors (mean[range]: 6[2–17] cm [63]. Tissue deformation may be induced by different daily stomach filling, tissue reaction to irradiation [167], and/or complications (e.g., mediastinitis in patient 13). The systematic and random errors of interfractional marker displacements were expected to be affected by the registration of the time-averaged CBCT with the “snapshot” pCT and the complication, radiation, or stomach filling-induced anatomical changes. Further, the less pronounced systematic and random errors in the LR and AP directions may be explained by the positions of the lungs, heart, and vertebrae with respect to the esophagus; their proximity can confine the motion of the esophagus in the LR and AP directions. The estimated margins in Table 2 indicate that, when markers are not available and bony anatomy-based registration is used, the internal margin required to account for the interfractional tumor position variation should always be large in the CC direction. Moreover, for tumors located at the GEJ (i.e., in the cardia region), the margin should also be large in the LR direction. However, due to the small sample size and the potential correlation between the marker displacements in one patient, the estimated margins might be. 3. 49. 527155-L-bw-Jin Processed on: 18-12-2018. PDF page: 49.

(13) Chapter 3. biased. Additionally, since the delineation errors, setup errors, and intrafractional uncertainties were not taken into account, this margin needs to be used with caution. Most previous studies were unable to investigate the tumor position variation in the CC direction due to the lack of fiducial markers [70, 116, 168–170] or quantified the displacement in the CC direction only for distal tumors near the GEJ with the aid of the GTV delineation, which is prone to observer variations [36]. Only one study quantified the interfractional position variation of esophageal tumors using metallic clips as markers identified in the 2D DRRs and follow-up onboard 2D kV images [159]. Despite reporting similar findings, its accuracy suffered from the use of 2D “snapshot” imaging data of each fraction. Moreover, metallic clips are more prone to detachment over time compared to the gold markers we used [61, 63, 159]. Further, they combined the data obtained for the distal esophagus and cardia regions, for which we observed a significant difference in marker displacements. The major limitation in our study is that the sample sizes in the proximal esophagus, middle esophagus, and cardia are small, which makes the comparisons for these subgroups sensitive to individuals. The marker-based registration was intended to improve setup verification and thus reduce the internal margins required to compensate for the interfractional tumor position variation. However, tissue deformation, evident from our analysis of pairwise distances between markers, led to the difficulties encountered in marker-based registration. For the 21 (9%) of 236 CBCT scans to which we successfully applied automatic registrations, the markers were clustered relatively closely together (pairwise distance: 12.7–63.4 mm) and none was located in the cardia into which over 80% of the esophageal tumors, i.e., distally located tumors, might extend. Moreover, the timeconsuming manual registration would make it unfeasible for clinical practice. A carina-based registration has been successfully implemented in lung cancer IGRT [171] and might be an alternative setup verification strategy in esophageal cancer IGRT as well. However, this has not yet been investigated. Consequently, bony anatomy registration is currently still the preferred method of setup verification in IGRT for esophageal cancer. Despite the fact that the use of markers for registration was not found to be clinically feasible, the use of markers could be of importance when dose escalation would be beneficial [49, 172]. Currently, to improve the IGRT for esophageal cancer, we recommend using markers and daily CBCT scans to inspect whether, after bony anatomy-based setup verification, the PTV covers the tumor volume adequately. Additionally, the daily CBCT allows inspection of whether other anatomical changes have taken place that could affect dosimetric parameters, as has been done in IGRT for lung cancer [173]. In conclusion, the position variation of esophageal tumors relative to bony anatomy is more pronounced in the CC direction and in the cardia. Although marker-based registration for setup verification is currently not feasible, the use of markers and daily CBCT can highly contribute to IGRT for esophageal cancer since they allow inspecting the PTV adequacy and anatomical changes after bony anatomy-based setup correction.. 50. 527155-L-bw-Jin Processed on: 18-12-2018. PDF page: 50.

(14)

No results found