Geometrical variability of esophageal tumors and its implications for accurate radiation therapy - Chapter 6: Quantification of respiration-induced esophageal tumor motion using fiducial markers and four-dimensional computed

(1)

UvA-DARE is a service provided by the library of the University of Amsterdam (https://dare.uva.nl)

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.

(2)

527155-L-bw-Jin 527155-L-bw-Jin 527155-L-bw-Jin 527155-L-bw-Jin Processed on: 18-12-2018 Processed on: 18-12-2018 Processed on: 18-12-2018

Processed on: 18-12-2018 PDF page: 85PDF page: 85PDF page: 85PDF page: 85

6

Quantification of respiration-induced esophageal

tumor motion using fiducial markers and

four-dimensional computed tomography

P. Jin, M.C.C.M. Hulshof, R. de Jong, J.E. van Hooft, A. Bel, and T. Alderliesten

A version of this chapter has been published in

Radiotherapy and Oncology. 2016; 118(3): 492–497

(3)

Abstract

Purpose

Respiration-induced tumor motion is an important geometrical uncertainty in esophageal cancer radiation therapy. The aim of this study was to quantify this motion using fiducial markers and four-dimensional computed tomography (4D-CT).

Materials and methods

Twenty esophageal cancer patients underwent endoscopy-guided marker implantation in the tu-mor volume and 4D-CT acquisition. The 4D-CT data were sorted into 10 breathing phases and the end-of-inhalation phase was selected as reference. We quantified for each visible marker (n = 60) the motion in each phase and derived the peak-to-peak motion amplitude throughout the breathing cycle. The motion was quantified and analyzed for four different regions and in three orthogonal directions.

Results

The median (interquartile range) of the peak-to-peak amplitudes of the respiration-induced marker motion (left–right/anterior–posterior/cranial–caudal) was 1.5(0.5)/1.6(0.5)/2.9(1.4) mm for the proximal esophagus (n = 6), 1.5(1.4)/1.4(1.3)/3.7(2.6) mm for the middle esophagus (n = 12), 2.6 (1.3)/3.3(1.8)/5.4(2.9) mm for the distal esophagus (n = 25), and 3.7(2.1)/5.3(1.8)/8.2 (3.1) mm for the cardia (n = 17).

Conclusions

The variations in the results between the three directions, four regions, and patients suggest the need of individualized region-dependent anisotropic internal margins. Therefore, we recommend using markers with 4D-CT to patient-specifically adapt the internal target volume (ITV). With-out 4D-CT, 3D-CTs at the end-of-inhalation and end-of-exhalation phases could be alternatively applied for ITV individualization.

(4)

666666

6.1 Introduction

The incidence of esophageal cancer has increased rapidly in the past decades [155,156]. Currently esophageal cancer is the eighth most common cancer worldwide [1]. Radiation therapy (RT) with concurrent chemotherapy has demonstrated benefits for patients with operable or inoperable esophageal cancer [17,20]. To generate the planning target volume (PTV) for RT of esophageal cancer, apart from the delineation uncertainty and interfractional tumor position variation, the uncertainty of respiration-induced tumor motion also needs to be taken into account. Although an active breathing control or a breath-holding technique could reduce this uncertainty, so far these techniques have been applied only in a few clinical RT trials for esophageal cancer [194,195]. Four-dimensional (4D) computed tomography (CT) was developed to facilitate the inspection of respiration-induced anatomical motion [196], however it has not yet commonly replaced the conventional “snapshot” three-dimensional (3D) CT for the treatment planning for esophageal cancer RT [17,197]. Therefore, it is necessary to quantify the respiration-induced tumor motion prior to incorporating this uncertainty into the internal margin [74].

With the aid of 4D-CT, most of the previous studies used gross tumor volume (GTV) delin-eation for the quantification of respiration-induced esophageal tumor motion [69,170,198,199]. However, without fiducial markers, the delineation of the primary tumor volume on CT may not be accurate even if18_{F-fluorodeoxyglucose-positron emission tomography is present [}₆₄_{]. Only}

one study placed large metal clips near the primary tumors as markers for motion quantification [165]. This was a rather small study, though, since only 12 patients with in total 22 markers were included.

Endoscopy/endoscopic ultrasound (EUS)-guided implantation of various types of small fidu-cial markers in the esophagus has recently been successfully performed [61–63] which aided the quantification of interfractional esophageal tumor position variation [176]. Consequently, in this retrospective study we included 20 esophageal cancer patients with in total 69 fiducial markers in the primary tumor volume. By the use of 4D-CT data and fiducial markers, we aimed to quan-tify the respiration-induced motion of the primary tumors located in different esophagus regions throughout the breathing cycle. In addition, we assessed the inter-observer variability in the mo-tion quantificamo-tion for method validamo-tion.

6.2 Materials and methods

Patients and markers

A previous prospective pilot study included 30 esophageal cancer patients who underwent endo-scopy/EUS-guided implantation of three different types of markers for evaluating and comparing the feasibility and benefits of marker implantation [63]. In this retrospective study, we included 20

(5)

Processed on: 18-12-2018 PDF page: 88PDF page: 88PDF page: 88PDF page: 88 patients with esophageal cancer between July 2013 and March 2015. Eighteen patients were from

the cohort of the pilot study who had given additional informed consent for 4D-CT acquisition; the other two patients were included after the pilot study ended. As described in [63], two differ-ent types of gold markers and one gel-based marker were used: solid marker (Cook Medical, Lim-erick, Ireland; or in-house manufactured), flexible coil-shaped marker (Visicoil; IBA Dosimetry, Bartlett, TN), and hydrogel marker (TraceIt; Augmenix Inc., Waltham, MA). For each patient, 2–6 markers of the same type were superficially placed in the submucosal layer at the cranial/caudal tu-mor borders and preferably the center of the primary tutu-mor. In total 69 markers were implanted. For gastroesophageal junctional tumors, the marker at the caudal border could be placed in the cardia or fundus of the stomach. The detailed procedure of marker implantation and the appear-ance of the three types of markers in the CT scans were described in [63]. An overview of patient and marker characteristics is presented in Table 6.1. According to the American Joint Commit-tee on Cancer staging manual [15], all markers were classified into four subgroups based on the marker locations in the 3D planning CT scans as done in [176]: the proximal esophagus, middle esophagus, distal esophagus, and cardia.

Table 6.1: Patient and marker characteristics.

Patient Sex Age Tumor_type Tumor_location Marker_type Markerlength/ volume Days between pCT and 4D-CT No. of markers at place-ment

No. of markers visible in 4D-CT Total Proximal Middle Distal Cardia

1 M 63 AC Distal Flexible 5–10 mm 0 4 4 0 0 2 2 2 F 65 SCC Middle Flexible 4 mm 42 4 3* 1 2 0 0 3 M 57 AC Distal Solid 5 mm 0 3 1*† 0 0 1 0 4 M 64 SCC Distal Flexible 2–10 mm 0 3 2* 0 0 1 1 5 F 67 SCC Middle Flexible 7–8 mm 45 4 4 1 3 0 0 6 M 71 SCC Distal Hydrogel 0.4 ml 0 6 5‡ 0 0 0 5 7 M 84 AC Distal Hydrogel 0.4 ml 0 3 3 0 1 2 0 8 M 61 AC Distal Flexible 10 mm 0 3 3 0 1 2 0 9 M 45 AC Distal Flexible 8 mm 0 3 3 0 0 3 0 10 M 63 AC Distal Flexible 8 mm 0 3 2* 0 0 2 0 11 F 59 AC Distal Flexible 8 mm 0 4 4 0 1 2 1 12 M 61 AC Distal Flexible 10 mm 0 3 2† 0 0 0 2 13 M 69 AC Distal Flexible 10 mm 0 4 4 0 0 2 2 14 M 69 AC Distal Flexible 10 mm 0 4 4 0 2 2 0 15 M 76 SCC Distal Hydrogel 0.4 ml 0 3 2‡ 0 0 2 0 16 M 65 SCC Middle Solid 5 mm 0 5 4* 1 1 1 1 17 M 75 SCC Distal Solid 5 mm 1 3 3 0 0 1 2 18 M 51 AC Full Solid 5 mm 0 3 3 1 1 1 0 19 M 73 SCC Proximal Flexible 10 mm 0 2 2 2 0 0 0 20 F 74 SCC Distal Flexible 10 mm 6 2 2 0 0 1 1 Total 69 60 6 12 25 17 *

Missing of marker was due to detachment of marker.

†_{Missing of marker was due to artifacts in 4D-CT reconstruction.}

‡_{Missing of marker was most likely due to absorption/dissolution of the hydrogel in the tissue.}

Diameter of gold markers: solid: 0.43–0.64 mm or 0.35–0.50 mm; flexible: 0.35 mm.

Abbreviations: M = male, F = female; AC = adenocarcinoma, SCC = squamous cell carcinoma; 4D-CT = four-dimensional computed tomography, pCT = planning computed tomography.

(6)

666666

4D-CT acquisition

For all patients, in addition to the 3D planning CT, a 4D-CT was acquired using a LightSpeed RT 16 CT scanner in cine mode (General Electric Company, Waukesha, WI). All 20 free-breathing patients were positioned supine with arms up above the head using an arm support (CIVCO Med-ical Solutions, Coralville, IA). No other immobilization devices were used. The axial thickness of the 4D-CT scan slices was 2.5 mm and the in-plane pixel size was 1.0 mm× 1.0 mm.

Based on the monitored breathing signal by the Real-time Position Management system (Var-ian Medical System, Palo Alto, CA), the 4D-CT data acquired in one breathing cycle were auto-matically sorted into 10 bins using the Advantage 4D software (General Electric). The image data in each bin were reconstructed into a 3D-CT scan, representing one breathing phase throughout the breathing cycle, where phase 0% denotes the end of inhalation and phase 50% approximately denotes the end of exhalation.

Marker identification and motion quantification

Using the X-ray Volume Imaging (XVI) software (version 4.5, Elekta AB, Stockholm, Sweden), the markers on each of the reconstructed 3D-CT scans were manually identified by two observers (R.d.J. and P.J.). The reconstructed 3D-CT scan of phase 0% (i.e., the end of inhalation) was se-lected as the reference scan. One trained radiation therapist (R.d.J.) manually registered for each patient the individual markers visible in each reconstructed 3D-CT scan of phases 10–90% to the corresponding markers in the reference scan. It was done by only altering the translations in XVI to align the centers of the two markers visually (example: Fig. 6.1). Based on the outcomes of the individual marker registrations, we calculated for each breathing phase the motion of each marker relative to its position in the reference. Then we derived for the individual markers the peak-to-peak amplitude of the respiration-induced motion that indicates the maximum marker position difference throughout the breathing cycle. All results were measured in the 3D vector distance, as well as in the left–right (LR), cranial–caudal (CC), and anterior–posterior (AP) directions, where the positive values indicate the left, cranial, and anterior direction, respectively.

Because of the elongated shape of the esophagus and the manner of motion of diaphragm and abdomen induced by respiration, the tumor motion could be direction- and location-dependent. Hence, we applied the Friedman test with Wilcoxon signed-rank test to compare the peak-to-peak amplitude of marker motion between the three orthogonal directions (LR, CC, and AP). Further, we applied the Kruskal–Wallis test with Dunn’s test to compare that between the four marker sub-groups (i.e., markers located in the proximal esophagus, middle esophagus, distal esophagus, and cardia). Holm adjustment was performed to all the post hoc tests. In this study, all the statistical analyses were performed using R software [144]. The significance level was set at 0.05.

(7)

Fig. 6.1: Example of the rigid marker-registration procedure (patient 1). The purple scan (in sagittal plane) represents

phase 0% (i.e., the end of inhalation) and the green scan represents phase 50% (i.e., approximately the end of

exha-lation).(a) The two scans were initially aligned on the vertebrae. (b) The marker (red arrow) was registered manually

by use of translations.

Inter-observer variability

To validate the marker-registration procedure and the quantification of the marker motion, the inter-observer variability was assessed. The second observer (P.J.) repeated all the marker regis-trations and derived the motion quantifications independently. For the three types of markers and all markers separately, the paired t-test was applied to compare the registration outcomes be-tween the two observers. Further, the one-way analysis of variance test was applied to compare the registration differences between the three types of markers.

6.3 Results

In total 60 out of 69 fiducial markers were identified: 6 in the proximal esophagus, 12 in the middle esophagus, 25 in the distal esophagus, and 17 in the cardia (Fig. 6.2). Due to the artifacts resulting from the reconstruction of the 3D-CT scans, detachments of solid and flexible coil-shaped mark-ers, or absorption/ dissolution of hydrogel spots, 9 markers could not be identified (Table 6.1). Both observers agreed on this identification result. No apparent deformations of hydrogel and flexible coil-shaped markers were observed due to the small dimension of the markers compared to the voxel size.

As shown in Fig. 6.3 and Fig. 6.4, we observed predominant marker motion in the CC direction throughout the breathing cycle for all four subgroups. However, for the four markers (two in the distal esophagus and two in the cardia) in patient 13, we observed evidently larger marker motion in the LR direction than in the CC and AP directions (Fig. 6.5).

The large interquartile range (IQR) and range of the peak-to-peak amplitudes in each subgroup indicate that the amplitude of the respiration-induced tumor motion was greatly different among the patients (Table 6.2 and Fig. 6.6). For all 60 markers, the median(IQR) of the peak-to-peak amplitudes of the respiration-induced marker motion was 2.4(1.9), 3.2(2.9), and 5.4(4.2) mm,

(8)

666666

Proximal esophagus Middle esophagus Distal esophagus Cardia n=6 n=12 n=25 n=17 Carina Inferior pulmonary vein Gastroesophageal junction

Fig. 6.2: Illustration of the locations of all 60 visible markers in the 20 patients. The ﬁlled circle denotes a marker.

Proximal esophagus Middle esophagus Distal esophagus Cardia n=6 n=12 n=25 n=17

Coronal View Sagittal View

Cranial Caudal Right Left Cranial Caudal Anterior Posterior 1 mm

Fig. 6.3: Average trajectories of the markers throughout the breathing cycle in the proximal esophagus, middle

esoph-agus, distal esophesoph-agus, and cardia. The trajectories are projected on the coronal and sagittal views of the esophagus. Note: the trajectories are not scaled to the illustration of the esophagus.

(9)

-5 0 5 10 15 Proximal esophagus (n = 6) -5 0 5 10 15 Middle esophagus (n = 12) -5 0 5 10 15 Distal esophagus (n = 25) -5 0 5 10 15 Cardia (n = 17) 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% -5 0 5 10 15 Overall (n = 60) 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% Left–Right Cranial–Caudal Anterior–Posterior

Three-dimensional vector distance

Breathing phase

Mean ± SD of

marker motion re

lative to breathin

g phase 0% [mm]

Fig. 6.4: Mean (symbol)_±standard deviations (SD, bar) of the marker motion throughout the breathing cycle for the markers located in the proximal esophagus, middle esophagus, distal esophagus, and cardia, as well as for all 60 markers (Overall). The motion was measured in the three directions, as well as the three-dimensional vector distance, where the positive values indicate the motion to the left, anterior, and cranial directions.

Marker 1 -5 0 5 10 15 LR CC AP 3D vector Marker 2 -5 0 5 10 15 Marker 3 -5 0 5 10 15 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% Marker 4 -5 0 5 10 15 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% Breathing phase

Marker motion relative to

breathing phase 0% [mm]

Fig. 6.5: Marker motion throughout a breathing cycle for the four markers located in the distal esophagus (markers

1 and 2) and cardia (markers 3 and 4) of patient 13. The motion was measured in the left–right (LR), cranial–caudal (CC), and anterior–posterior (AP) directions, as well as the three-dimensional vector distance (3D vector), where the positive values indicate the motion to the left, anterior, and cranial directions.

(10)

666666

in the LR, CC, and AP direction, respectively. The peak-to-peak amplitude in the CC direction was significantly larger than that in the LR and AP directions (p<0.05) for markers located in the middle esophagus, distal esophagus, and cardia, but not for markers in the proximal esophagus. Further, the 3D vector peak-to-peak motion amplitude (median[IQR]) for the markers located in the distal esophagus (7.7[2.4] mm, n = 25) and cardia (10.0[2.7] mm, n = 17) was significantly larger than that in the proximal esophagus (3.1[1.6] mm, n = 6) and middle esophagus (3.9[2.8], n = 12) (p<0.05).We observed the same findings for the peak-to-peak amplitude in each of the three orthogonal directions. Only for the peak-to-peak motion amplitude in the AP direction and in the 3D vector distance, a significant difference also existed between the markers located in the distal esophagus and cardia (p<0.05) (Table 6.2 and Fig. 6.6).

Table 6.2: Median, interquartile range (IQR), and range of the peak-to-peak respiration-induced motion amplitude

in the left–right (LR), cranial–caudal (CC), and anterior–posterior (AP) directions, for markers located in the proximal esophagus, middle esophagus, distal esophagus, and cardia. Note: n is the number of markers.

n Median [mm] IQR [mm] Range [mm]

LR CC AP LR CC AP LR CC AP

Proximal esophagus 6 1.5 2.9 1.6 0.5 1.4 0.5 1.0 – 2.4 1.4 – 4.0 1.2 – 3.0

Middle esophagus 12 1.5 3.7 1.4 1.4 2.6 1.3 0.0 – 2.7 0.3 – 8.2 0.0 – 3.2

Distal esophagus 25 2.6 5.4 3.3 1.3 2.9 1.8 1.6 – 10.0 3.7 – 11.7 1.3 – 7.0

Cardia 17 3.7 8.2 5.3 2.1 3.1 1.8 1.2 – 11.0 2.8 – 14.0 2.3 – 8.1

For the flexible coil-shaped markers, the marker registrations of the two observers were merely significantly different in the LR direction (p = 0.002), but not in the AP and CC directions (Table 6.3). For the solid and hydrogel markers, however, no significant difference in the marker registra-tions between the two observers was manifested in all three direcregistra-tions. Moreover, for the marker registration results of the two observers in the three directions, we found no significant differences between the three types of markers.

Table 6.3: Mean, standard deviation (SD), and range of the diﬀerences in the marker registrations in the left–right

(LR), cranial–caudal (CC), and anterior–posterior (AP) directions between the two observers, for solid markers, ﬂexible coil-shaped markers, hydrogel markers, and all markers (overall). Note: n is the number of registrations, which equals the number of markers multiplied by nine.

n Median [mm] IQR [mm] Range [mm]

LR CC AP LR CC AP LR CC AP Solid 99 −0.1 −0.1 0.0 0.8 1.0 1.0 −2.8 – 2.4 −3.2 – 2.6 −3.2 – 3.4 Flexible 351 0.1 0.0 −0.1 0.9 1.0 1.0 −3.3 – 2.9 −3.3 – 3.0 −4.1 – 3.8 Hydrogel 90 0.1 0.1 0.1 0.5 1.0 0.6 −1.6 – 1.6 −2.9 – 3.8 −1.6 – 2.0 Overall 540 0.1 0.0 0.0 0.8 1.0 0.9 −3.3 – 2.9 −3.3 – 2.8 −4.1 – 3.8

93

(11)

0 5 10 15 20 Proximal esophagus (n = 6) 0 5 10 15 20 Middle esophagus (n = 12) 0 5 10 15 20 Distal esophagus (n = 25) 0 5 10 15 20 Cardia (n = 17) LR CC AP 3D 0 5 10 15 20 Overall (n = 60) LR CC AP 3D Boxes: Whiskers:

Upper and lower quartiles

Highest/lowest data within 1.5 × interquartile range of the upper/lower quartiles Median

Outlier Data point

Direction

Peak-to-peak amplitude of

marker motion over the breathi

ng cycle [mm]

Fig. 6.6: Distribution of the peak-to-peak amplitude of the marker motion throughout the breathing cycle for the

markers located in the proximal esophagus, middle esophagus, distal esophagus, and cardia, as well as for all markers (Overall). The motion was measured in the left–right (LR), cranial–caudal (CC), and anterior–posterior (AP) directions, as well as the three-dimensional vector distance (3D).

6.4 Discussion

We quantified the respiration-induced motion of esophageal tumors in detail by the use of 60 fidu-cial markers and 4D-CT. We found the motion amplitude to be dependent on its direction and the tumor location.

The average trajectories of the markers depicted in Fig. 6.3 imply that the motion was predomi-nantly induced by respiration since the diaphragm moves along the CC direction and the chest and the belly move along the AP direction during respiration. Although the cardiac-induced motion cannot be removed from our results, it is expected to be far smaller than the respiration-induced motion [121].

Previous studies showed that the esophageal tumor motion between the end of inhalation and the end of exhalation is the largest in the CC direction, where the mean± SD of the amplitudes measured with the GTV delineations was 3.9± 2.7 (LR), 8.7 ± 4.7 mm (CC), and 3.8 ± 2.3 (AP) for gastroesophageal junction tumors (n = 25) [11]; 2.8± 2.0 (LR), 8.0 ± 4.5 mm (CC), and 2.2

± 2.3 (AP) for thoracic esophageal tumors (n = 35) [199]. For esophageal tumors located in the abdomen, the respiration-induced motion was found to be larger than that in the thorax (n = 31) [69]. However, due to the potential large intra-observer variation of GTV delineation (median

(12)

666666

SD: 5.3 mm reported in [200], the measurement accuracy could be questionable even though the GTV centroid was used. Similar to our study, one study, including 12 patients, investigated the motion of in total 22 metal clips near the tumor volume in the three orthogonal directions for the proximal, middle, and distal esophagus [165]. However, based on our experience, the use of metal clips could introduce more imaging artifacts due to the 3–4 times larger diameter than those of the gold markers that we used. This could thereby affect the measurement by means of a deviation in the marker center determination using the semiautomatic approach given in [165] or a larger inter-observer variation using our quantification approach. Further, this study gave no indications of the motion in the region below the diaphragm where we found significantly larger peak-to-peak motion amplitudes in the AP direction than in the thoracic region (p<0.05). Moreover, based on our observations, the two phases used for deriving the peak-to-peak amplitude of the motion in the CC direction were not always the same as those for the LR direction (e.g., Fig. 6.5). For the previous studies that only selected the scans of end-of-inhalation and end-of-exhalation phases, the quantification of the peak-to-peak motion amplitude would therefore be questionable [69,

199].

In most cases, the inter-observer variation in marker registration was very small compared to the marker motion, even though a significant difference (mean: 0.1 mm) was found for the flexi-ble coil-shaped markers in the LR direction (Taflexi-ble 6.3). The difference was mainly caused by the varied shapes of marker artifacts in the reconstructed 3D-CT scans between different phases. Fur-ther, the statistical analysis indicated that the marker registration results were unlikely influenced by the use of three different types of markers. Additionally, the estimated errors caused by the limited imaging resolution should be within half of a voxel size, which are inevitable [201]. There-fore, the marker motion quantification ought to be valid, as long as these markers did not migrate during the 4D-CT acquisition, which should be the case.

In addition to the quantification of respiration-induced motion of esophageal tumors in pre-vious studies, the margins required to compensate for this motion were proposed in case when 4D-CT was not available for treatment planning [68,69,165,170,199]. Although these pro-posed margins were able to account for 95% of the respiration-induced motion in their popula-tions, they might be too large to be directly incorporated into the PTV. First, our study indicates a large variation in respiration-induced motion among patients (Table 6.2 and Fig. 6.6). Further, the respiration-induced tumor motion in the CC direction might not necessarily be predominant (e.g., Fig. 6.5). Moreover, the 3D-CT used for treatment planning cannot indicate in which breath-ing phase the patient was durbreath-ing CT acquisition. Hence, these proposed margins based on the peak-to-peak motion amplitude might be an overestimation, resulting in a potential toxicity of the organs at risk. In summary, we assert that it might be inappropriate to propose general margins required to compensate for the respiration-induced tumor motion.

4D-CT has been commonly used in the treatment planning for lung cancer and pancreatic can-cer RT and the individualized internal target volume (ITV) and mid-ventilation concepts have

(13)

Processed on: 18-12-2018 PDF page: 96PDF page: 96PDF page: 96PDF page: 96 been introduced and widely discussed [127,202]. Meanwhile, the use of markers has revealed

its advantages in GTV delineation and daily image-guided setup verification in RT for esophageal cancer [61,63,176]. Therefore, instead of using a 3D-CT with an isotropic clinical target volume (CTV)-PTV margin as is common practice in most clinical protocols (e.g., [17]), we recommend using 4D-CT with (e.g., flexible coil-shaped) markers to improve the GTV delineation, individu-alize the ITV with anisotropic internal margins, and adapt the CTV-PTV margins accordingly in the RT treatment planning for esophageal cancer. In combination with the proposal in [176], the daily setup verification would also be influenced by the use of 4D-CT. When 4D-CT acquisition is not possible, the use of 3D-CT scans at both end-of-inhalation and end-of-exhalation phases might be the alternative, despite the fact that the two breath-holding phases could overestimate the peak-to-peak amplitude during the treatment course.

The limitation in our study lies in the small data population of the markers located in the prox-imal esophagus (6 markers), which requires more data for a solid conclusion. However, the am-plitude of marker motion in the proximal esophagus is still expected to be smaller than that in the other three regions with more data included. This would therefore possibly still yield the anisotropic and region-dependent motion. Further, the interfractional variation of respiration-induced esophageal tumor motion has not yet been investigated. Such a study has previously been done for pancreatic tumors [203]. With the markers being visible also in the follow-up cone-beam CT (CBCT) scans, it may be feasible to also investigate this for the included esophageal cancer patients in our study [176]. If the respiration-induced tumor motion observed in the pre-treatment 4D-CT is found not representative for that during the pre-treatment course and meanwhile 4D-CBCTs are available for tumor-/marker-based setup verification, using the mid-ventilation ap-proach as done in RT for lung cancer would be a possibility [128]. Otherwise, an active breathing control or a breath-holding technique could be applied as [194,195] did. These will be investi-gated in the follow-up studies.

In conclusion, the respiration-induced motion was found to be the largest in the CC direc-tion and for the distally located esophageal tumors, suggesting anisotropic internal margins are re-quired to take induced motion into account. Due to the large variation in respiration-induced motion among patients, we recommend using the combination of markers and 4D-CT or 3D-CTs acquired at both end-of-inhalation and end-of-exhalation phases for ITV individualiza-tion. However, if the interfractional respiration-induced motion variation is large, the feasibility of a mid-ventilation approach, an active breathing control technique, or a breath-holding technique should be investigated alternatively.