Geometrical variability of esophageal tumors and its implications for accurate radiation therapy - Chapter 8: Interfractional variability of respiration-induced esophageal tumor motion quantified using fiducial markers and

(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: 117PDF page: 117PDF page: 117PDF page: 117

8

Interfractional variability of respiration-induced

esophageal tumor motion quantified using fiducial

markers and four-dimensional cone-beam

computed tomography

P. Jin, M.C.C.M. Hulshof, N. van Wieringen, A. Bel, and T. Alderliesten

A version of this chapter has been published in

Radiotherapy and Oncology. 2017; 124(1): 147–454

(3)

Abstract

Purpose

To investigate the interfractional variability of respiration-induced esophageal tumor motion using fiducial markers and four-dimensional cone-beam computed tomography (4D-CBCT) and assess if a 4D-CT is sufficient for predicting the motion during the treatment.

Materials and methods

Twenty-four patients with 63 markers visible in the retrospectively reconstructed 4D-CBCTs were included. For each marker, we calculated the amplitude and trajectory of the respiration-induced motion. Possible time trends of the amplitude over the treatment course and the interfractional variability of amplitudes and trajectory shapes were assessed. Further, the amplitudes measured in the 4D-CT were compared to those in the 4D-CBCTs.

Results

The amplitude was largest in the cranial–caudal direction of the distal esophagus (mean: 7.1 mm) and cardia (mean: 7.8 mm). No time trend was observed in the amplitude over the treatment course. The interfractional variability of amplitudes and trajectory shapes was limited (mean: ⩽1.4 mm). Moreover, small and insignificant deviation was found between the amplitudes quan-tified in the 4D-CT and in the 4D-CBCT (mean absolute difference:⩽1.0 mm).

Conclusions

The limited interfractional variability of amplitudes and trajectory shapes and small amplitude difference between 4D-CT-based and 4D-CBCT-based measurements imply that a single 4D-CT would be sufficient for predicting the respiration-induced esophageal tumor motion during the treatment course.

(4)

88888888

8.1 Introduction

Combined with concurrent chemotherapy, radiation therapy (RT) has become standard as part of neoadjuvant or definitive therapy for esophageal cancer [17,18]. How to cope with the uncer-tainties in tumor delineation, interfractional tumor position variation, and intrafractional tumor motion such as respiration-induced motion, is one concern for RT of esophageal cancer [134]. With the evolution from three-dimensional (3D) conformal RT to intensity-modulated RT and volumetric-modulated arc therapy, with which a more conformal dose distribution can be ob-tained [75,219,220], it becomes more crucial to deal with these uncertainties to ensure accurate delivery of the dose to the target volume while sparing the organs at risk as much as possible [221]. Respiration-induced esophageal tumor motion is one of the major sources of intrafractional uncertainties. The quantification of this motion was, until recently, mainly done by measuring the displacement of the delineated gross tumor volume or anatomical landmarks in the 4D computed tomography (4D-CT) data, in spite of the limited soft-tissue contrast of the 4D-CT [69,170,198]. Since the endoscopy-/endoscopic ultrasound (EUS)-guided implantation of fiducial markers in the volume of esophageal tumor was successful and no migration of fiducial markers was found during the treatment course [63,176], the quantification of the respiration-induced esophageal tumor motion using fiducial markers and 4D-CT became more accurate [165,210]. In these studies, however, the respiration-induced esophageal tumor motion was measured only within one 4D-CT acquisition per patient and the interfractional variability of the respiration-induced motion has not yet been investigated.

Apart from 4D-CT, 4D cone-beam CT (CBCT) can be used for quantifying the respiration-induced tumor motion, as done previously for lung tumors [67,212]. This allows the quantifica-tion of the daily respiraquantifica-tion-induced tumor moquantifica-tion. However, 4D-CBCT has not yet been com-monly introduced in esophageal cancer RT. Recently, a phantom study compared the visibility of gold markers in the 4D-CBCTs acquired using multiple settings by altering the dose, gantry rota-tion speed, and fluoroscopy projecrota-tion image number. It has demonstrated that fiducial markers can be sufficiently visible in the 4D-CBCTs reconstructed using the same fluoroscopic projection images as used for a clinical 3D-CBCT reconstruction [222]. High accuracy of using the fiducial markers to manually quantify the respiration-induced motion was also shown regardless of the 4D-CBCT acquisition settings.

The aim of this study was to investigate the interfractional variability of respiration-induced esophageal tumor motion using fiducial markers and 4D-CBCT scans which were retrospectively reconstructed using the projection images of the clinical 3D-CBCT scans. Using these data, we verified whether a single measurement based on the 4D-CT acquisition [210] is a good predicator for the respiration-induced tumor motion during the treatment course (i.e., has a small difference from the measurements based on the 4D-CBCT).

(5)

8.2 Materials and methods

Patients and markers

We included 24 esophageal cancer patients with implanted gold markers, who were also included in former studies [63,176,210]. Two different types of gold markers were implanted: a solid marker (Cook Medical, Limerick, Ireland; or in-house manufactured) and a flexible coil-shaped marker (Visicoil; IBA Dosimetry, Bartlett, TN) [63]. For each patient, 2–5 markers were placed at the cranial and caudal border of the primary tumor and, preferably, in the center of the tumor. The details of the patients and markers are listed in Table 8.1. In total, 63 markers remained visible in the reconstructed 4D-CBCTs over the whole treatment course. They were categorized accord-ing to the American Joint Committee on Cancer manual [15] into four subgroups based on the region of the esophagus in which the marker was located: proximal esophagus (n = 13), mid-dle esophagus (n = 10), distal esophagus (n = 28), and cardia (n = 12), as previously illustrated [176,210].

Table 8.1: Overview of patient and marker characteristics.

Patient Sex Age Tumor_type Tumor_location Marker_type

Marker length [mm] Dose scheme [Gy] No. of CBCTs No. of markers At placement Visible in pCT Visible in CBCTs 1 M 56 AC Distal Solid 5 23 × 1.8 7 2 1 1 2 M 79 AC Distal Solid 5 23 × 1.8 7 3 2 1 3 M 62 AC Distal Solid 5 23 × 1.8 7 3 3 1 4 M 61 PDC Distal Solid 5 23 × 1.8 7 3 3 3→ 2 5 F 57 SCC Distal Solid 5 28 × 1.8 8 4 4 4 6* M 63 AC Distal Flexible 5–10 23 × 1.8 8 4 4 4 7* F 65 SCC Middle Flexible 4 23 × 1.8 8 4 3 3 8* M 64 SCC Distal Flexible 2–10 23 × 1.8 12 3 2 2 9 M 70 SCC Proximal Flexible 10 28 × 1.8 8 3 3 3 10* F 67 SCC Middle Flexible 7–8 23 × 1.8 23 4 4 4 11* M 61 AC Distal Flexible 10 23 × 1.8 8 3 3 3 12* M 45 AC Distal Flexible 8 23 × 1.8 7 3 3 3→ 2 13* M 63 AC Distal Flexible 8 23 × 1.8 11 3 2 2 14 F 79 AC Distal Flexible 8 28 × 1.8 8 4 4 3 15* F 59 AC Distal Flexible 8 23 × 1.8 9 4 4 4 16* M 61 AC Distal Flexible 10 23 × 1.8 12 3 3 3 17* M 69 AC Distal Flexible 10 23 × 1.8 12 4 4 2 18* M 69 AC Distal Flexible 10 23 × 1.8 8 4 4 4 19 F 67 SCC Proximal Solid 5 28 × 1.8 8 3 3 3 20* M 65 SCC Middle Solid 5 28 × 1.8 9 5 4 4 21* M 75 SCC Distal Solid 5 23 × 1.8 9 3 2 2 22* M 51 AC Full Solid 5 23 × 1.8 7 3 3 2 23* M 73 SCC Proximal Flexible 10 28 × 2.2 28 2 2 2 24* F 74 SCC Distal Flexible 10 23 × 1.8 23 2 2 2 Total 254 79 72 65→ 63

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 = plan-ning computed tomography, CBCT = cone-beam computed tomography.

*

Patients included for comparison between 4D-CT and 4D-CBCT data.

Within eight days (median: one day) after marker implantation, a 3D planning-CT (pCT) was acquired for each patient (LightSpeed RT 16 CT; General Electric, Waukesha, WI). All patients

(6)

88888888

underwent the scanning in head-first supine position with an arm and knee support (CIVCO Med-ical Solution, Coralville, IA). The thickness of the CT slices was 2.5 or 3.0 mm and the in-plane pixel size was 1.0, 1.2, or 1.3 mm depending on the field of view (FOV) of the scan. For 16 out of the 24 patients, who were previously included in [210], 4D-CT was acquired in addition to the pCT (Table 8.1). The 4D-CT acquisition details can be found in [210].

Prior to the treatment fractions, 3D-CBCT scans were routinely acquired following an online setup verification protocol (for patients 10, 23, and 24) or an extended no action level (eNAL) setup verification protocol [114] (for the remainder of the patients) based on bony anatomy (i.e., vertebrae). When the eNAL protocol was used, daily 3D-CBCTs were acquired for the first four fractions, followed by once-weekly acquisitions. Additional 3D-CBCT scans were acquired when the setup correction exceeded the tolerance in the NAL phase. Consequently, 7–28 (median: 8) 3D-CBCT scans were obtained for each patient. During pCT/3D-CBCT scanning and treatment, all patients were freely breathing without receiving any training, coaching or feedback related to achieving a stable breathing pattern.

4D-CBCT reconstruction

4D-CBCTs were reconstructed retrospectively using the available fluoroscopic projection images of the clinical 3D-CBCT scans. These clinical 3D-CBCT scans were acquired using the CBCT imaging units mounted on linear accelerators (Synergy; Elekta AB, Stockholm, Sweden). Ap-proximately 660 fluoroscopy projection images were collected over a full arc of 360° with a shifted detector, yielding a medium FOV of 410× 410 mm2_{in the axial plane.}

The breathing signal was automatically extracted by detecting the position of diaphragm-like features in the projection images based on the so-called Amsterdam Shroud algorithm (MAT-LAB 2013b, The MathWorks Inc., Natick, MA) [217]. All projection images were then sorted and reconstructed in 10 breathing phases (0–90%) to obtain the 4D-CBCT (i.e., 10 3D-breathing-phase scans) with an isotropic 1.0 mm voxel size using X-ray Volume Imaging software (XVI 4.5.0; Elekta) (Fig. 8.1).

Respiration-induced motion quantification

First, each 3D-breathing-phase scan of the 4D-CBCT was rigidly registered to the pCT based on bony anatomy, i.e., vertebrae, in XVI (Fig. 8.1). Using the coordinate system of the pCT as a ref-erence, marker positions in the 4D-CBCT relative to those in the pCT were then obtained by manually registering each 3D-breathing-phase scan to the pCT with respect to the marker cen-troid using translations only. This was done for each marker separately.

For each 4D-CBCT, the motion trajectory of each marker over the respiration cycle was sub-sequently depicted by assessing the differences of the 10 marker positions relative to the

(7)

Processed on: 18-12-2018 PDF page: 122PDF page: 122PDF page: 122PDF page: 122 Fig. 8.1: Example (patient 15) of four-dimensional cone-beam computed tomography at the end of inhalation (top row)

and end of exhalation (bottom row) breathing phases (green) overlaid on the three-dimensional planning computed tomography (purple) with both scans registered on the vertebrae. The arrow indicates the position of a marker.

tory centroid, i.e., mean marker position. The respiration-induced motion was quantified in the left–right (LR), cranial–caudal (CC), and anterior–posterior (AP) directions, with the positive coordinates to the left, cranial, and anterior. The peak-to-peak amplitude of respiration-induced motion (hereafter referred to as amplitude) was calculated as the maximum position difference of the motion in each direction.

Per marker, the mean and standard deviation (SD) of the quantified amplitudes over all the 4D-CBCT scans were calculated as the representations of the interfractional mean amplitude and interfractional variability of amplitudes, respectively. The interfractional minimum and maximum amplitudes and the range, i.e., difference between maximum and minimum, were also calculated. Moreover, per marker, the SD of the positions at the end of inhalation relative to the trajectory centroid over all 4D-CBCT scans was calculated. The SD of the positions at the end of exhalation relative to the trajectory centroid was also calculated. Both the SDs were considered as measures of the interfractional variability of trajectory shapes since the end of inhalation and the end of exhalation phases are the two extreme breathing phases, which dominantly define the trajectory shape of the respiration-induced tumor motion. Fig. 8.2 illustrates how these were calculated.

Statistical analysis

Linear mixed-effects models were used in our analyses to account for the intra-patient correlation of the amplitudes due to the different number of markers among the patients by taking the patient identification number as a random effect in all the tested models. Because of the unequal number of CBCTs among the patients, the marker identification number was also taken as a random effect in the models [145].

(8)

88888888

4D-CBCT #1

4D-CBCT #2

!"!"

Marker positions relative to the trajectory centroid Amplitudes (Left-Right, Cranial-Caudal, Anterior-Posterior)

Left Right Anterior Posterior Caudal Cranial

Marker position in each breathing phase (0–90%) Trajectory centroid

Peak-to-peak amplitudes of the respiration-induced motion in three directions 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% Last 4D-CBCT

Marker positions relative to the trajectory centroid Amplitudes (Left-Right, Cranial-Caudal, Anterior-Posterior)

#1 #2 Last

!

Am

plitude

4D-CBCT

Interfractional mean amplitude

Interfractional minimum, maximum, and range of amplitudes

Interfractional variability of amplitudes, i.e., standard deviation (SD)

Left Right Anterior Posterior Caudal Cranial Interfractional variability of trajectory shapes: Exhalation Interfractional variability of trajectory shapes: Inhalation

0% 50% Marker position in

each 4D-CBCT Trajectory centroid Average marker position relative to centroid

SD sphere of the marker positions relative to centroid

}

Fig. 8.2: Schematic illustration of calculation of the interfractional mean, variability, minimum, maximum, and range of

amplitudes and the interfractional variability of trajectory shapes, i.e., marker position variation relative to the trajectory centroid at the end of inhalation and exhalation phases for each marker. Abbreviation: 4D-CBCT = four-dimensional cone-beam computed tomography.

By taking the number of days after starting treatment, which was derived from the date of CBCT acquisition, as a fixed effect, we investigated whether the amplitudes have a possible time trend over the treatment course. In addition, when taking the interfractional mean amplitude as a fixed effect, we tested whether the interfractional variability of amplitudes and trajectory shapes (i.e., the aforementioned SDs) was linearly correlated with the interfractional mean amplitude by cal-culating the marginal and conditional coefficients of determination (R2) [223]. Moreover, the in-terfractional mean amplitude and variability of amplitudes and trajectory shapes were compared between the LR, CC, and AP directions when the direction was taken as a fixed effect. Further, with stratifying the direction and taking the region of the esophagus as a fixed effect, they were compared between the four regions: proximal, middle, distal esophagus, and cardia.

To assess the representativeness of the respiration-induced motion measured in a single pre-treatment 4D-CT acquisition for that during the pre-treatment course, we assessed if the amplitudes measured on the 4D-CBCT scans were different from those derived from the 4D-CT scans by taking image modality (i.e., 4D-CT or 4D-CBCT) as a fixed effect. This comparison was done only for 44 markers in 16 patients who had 4D-CT acquisitions with visible markers and were included in the previous study [210].

For testing the linearity, homoscedasticity, and normality of these models, the residuals of the data were examined afterward. The significance level was set to 0.05. All statistical tests in this study were done using R software [144].

(9)

8.3 Results

For over 97% of the markers, the interfractional mean amplitude was <10.0 mm in all three direc-tions. The average motion trajectory and its 95% confidence interval (CI) in the four regions of the esophagus illustrated in Fig. 8.3 indicate a predominant respiration-induced tumor motion in the CC direction and in the distal esophagus and cardia. For the three directions and four regions, Fig. 8.4 illustrates the amplitudes for individual markers and Table 8.2 lists the interfractional mean, variability, minimum, maximum, and range of amplitudes as well as the interfractional variability of trajectory shapes. n=13 n=10 n=28 n=12 Cardia

Coronal View Sagittal View

Cranial Caudal Right Left Cranial Caudal Anterior Posterior 1.0 mm Proximal esophagus Middle esophagus Distal esophagus

Fig. 8.3: Illustration of the average trajectory (black lines) and its 95% conﬁdence interval (gray area) of the marker

motion throughout the breathing cycle in the four regions. The trajectories are projected on the coronal (left) and sagittal (right) views of the schematic esophagus drawing. Note: the trajectories and the 1.0 mm scale are not scaled to the esophagus drawing.

The interfractional variability of amplitudes was small (mean: ⩽1.4 mm, Table 8.2) for the three directions and four regions. Further, regardless of the direction and region, the SDs of the marker positions relative to the trajectory centroid at the end of inhalation and exhalation phases were tiny (mean:⩽1.0 mm) and they were even slightly smaller at the exhalation than at the in-halation (Table 8.2). These imply that the interfractional variability of trajectory shapes was lim-ited and the markers positions at the exhalation phase were slightly more stable than at the inhala-tion phase.

By both visual inspections of the amplitudes over time (examples: Fig. 8.5) and applying the

(10)

88888888

Left–Right 0 5 10 15 20 25 Proximal esophagus Middle esophagus Distal esophagus Cardia

Interfractional mean amplitude

Cranial–Caudal 0 5 10 15 20 25 Anterior–Posterior 0 5 10 15 20 25 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Patient number Peak-to-peak amplitude [m m]

Fig. 8.4: Amplitude of the respiration-induced motion over the treatment course for all the 63 markers in 24 patients

in the left–right, cranial–caudal, and anterior–posterior directions.

Table 8.2: Mean (95% conﬁdence interval [CI]) of the interfractional mean amplitude, interfractional variability of

am-plitudes and trajectory shapes (i.e., marker position variation relative to the trajectory centroid at the end of inhalation and exhalation breathing phases), interfractional minimum and maximum of amplitudes, and range (i.e., the diﬀerence between the maximum and minimum) in the left–right (LR), cranial–caudal (CC), and anterior–posterior (AP) directions for the four regions. Note: n is the number of markers.

Interfractional mean amplitude Interfractional variability of amplitudes Interfractional variability of

trajectory shapes Interfractionalminimum amplitude Interfractional maximum amplitude Interfractional range of amplitudes Inhalation Exhalation Proximal esophagus (n = 13) LR 1.8(0.9–2.7) 0.5(0.3–0.6) 0.4(0.2–0.5) 0.3(0.2–0.4) 1.2(0.4–1.9) 2.5(1.3–3.8) 1.5(0.9–2.0) CC 4.4(3.1–5.8) 0.8(0.5–1.0) 0.5(0.3–0.7) 0.5(0.3–0.7) 3.4(2.0–4.9) 5.9(4.1–7.7) 2.6(1.7–3.5) AP 1.4(0.8–2.0) 0.5(0.3–0.6) 0.4(0.2–0.6) 0.3(0.2–0.4) 0.8(0.2–1.4) 2.3(1.5–3.1) 1.5(1.0–2.1) Middle esophagus (n = 10) LR 1.9(1.0–2.7) 0.5(0.4–0.7) 0.4(0.2–0.6) 0.4(0.3–0.5) 1.0(0.3–1.8) 2.7(1.5–3.8) 1.8(1.2–2.4) CC 4.9(3.6–6.3) 0.8(0.6–1.1) 0.6(0.4–0.8) 0.5(0.3–0.7) 3.7(2.3–5.1) 6.5(4.8–8.3) 2.9(2.0–3.7) AP 1.6(1.0–2.2) 0.6(0.4–0.8) 0.5(0.3–0.7) 0.4(0.3–0.5) 1.0(0.4–1.5) 3.2(2.3–4.1) 2.2(1.6–2.8) Distal esophagus (n = 28) LR 3.0(2.2–3.8) 0.7(0.6–0.8) 0.6(0.5–0.7) 0.5(0.4–0.6) 2.1(1.4–2.7) 4.2(3.3–5.2) 2.2(1.8–2.5) CC 7.1(5.9–8.3) 1.1(0.9–1.3) 0.9(0.8–1.0) 0.7(0.6–0.8) 5.3(4.0–6.5) 8.8(7.3–10.3) 3.5(2.9–4.2) AP 3.0(2.6–3.4) 0.8(0.7–1.0) 0.7(0.5–0.8) 0.5(0.5–0.6) 1.9(1.5–2.3) 4.4(3.9–5.0) 2.5(2.2–2.9) Cardia (n = 12) LR 3.7(2.8–4.6) 0.9(0.8–1.1) 0.7(0.5–0.9) 0.6(0.5–0.7) 2.4(1.6–3.1) 5.6(4.4–6.8) 2.9(2.4–3.5) CC 7.8(6.5–9.1) 1.4(1.2–1.7) 1.0(0.8–1.3) 0.8(0.6–0.9) 5.7(4.3–7.1) 9.9(8.2–11.7) 4.2(3.4–5.1) AP 4.6(3.9–5.3) 1.0(0.8–1.2) 1.0(0.8–1.2) 0.6(0.5–0.7) 3.1(2.5–3.7) 5.9(5.0–6.9) 3.1(2.5–3.7) Unit: mm

linear mixed-effects model, no obvious time trend over the treatment period of 4.5–5.5 weeks can be generalized for the amplitudes irrespective of the region, with the absolute slope <0.01 (SD<0.05) mm/day in all three directions. Despite the significant difference from 0 mm/day in slope in the LR direction (p = 0.005), the mean slope was only 0.008 mm/day. No significant difference in slope was found for the CC and AP directions.

Both the marginal R2_{and conditional R}2_{indicate that the interfractional variability of}

ampli-tudes and trajectory shapes was well explained by the model, when taking both the fixed and

(11)

0 2 4 6 8 10 12 Patient 6 (4 markers) Proximal esophagus Middle esophagus Distal esophagus Cardia 0 2 4 6 8 10 12 0 2 4 6 8 10 12 1 5 10 15 20 25 30 35 0 2 4 6 8 10 12 Patient 15 (4 markers) 0 2 4 6 8 10 12 0 2 4 6 8 10 12 1 5 10 15 20 25 30 35 0 2 4 6 8 10 12 Left–Right Patient 24 (2 markers) 0 2 4 6 8 10 12 Cranial–Caudal 0 2 4 6 8 10 12 Anterior–Posterior 1 5 10 15 20 25 30 35

Day number counting from the first treatment day

Amplitude [mm]

Fig. 8.5: Amplitude of the respiration-induced motion over the treatment course for 10 markers in three patients

(patients 6, 15, and 24) in the left–right, cranial–caudal, and anterior–posterior directions.

dom effects into consideration (Fig. 8.6). All the slopes of fits being significantly larger than 0 implies that the interfractional variability of amplitudes and trajectory shapes was positively cor-related with the interfractional mean amplitude in all three directions. Further, the somewhat smaller slopes of the fits for the SDs of the marker positions relative to the trajectory centroid at the exhalation phase compared to those at the inhalation phase suggest that the interfractional mean amplitude had less influence on the interfractional variability of trajectory at the exhalation phase than the inhalation phase (Fig. 8.6).

According to Table 8.2, regardless of the region, the interfractional mean amplitude in the CC direction was always significantly larger than that in the LR and AP directions (p<0.001). In the cardia, a significant difference was observed in the interfractional mean amplitude between the LR and AP directions (p<0.001). Moreover, irrespective of the direction, both the interfractional mean amplitudes in the distal esophagus and cardia were significantly larger than that in the prox-imal and middle esophagus (p<0.001).

Only in the proximal and distal esophagus, a significantly larger interfractional variability of amplitudes and trajectory shapes was observed in the CC direction than that in the LR and AP di-rections (p<0.01). Additionally, in the CC direction, the interfractional variability of amplitudes and trajectory shapes in both the distal esophagus and cardia was significantly larger compared to that in the proximal and middle esophagus (p<0.01). In the LR and AP directions, the

(12)

88888888

Slope: 0.12 *** Marginal R2_{= 0.57} Conditional R2_{= 0.65} Proximal esophagus Middle esophagus Distal esophagus Cardia Amplitudes 0 1 2 3 Slope: 0.10 *** Marginal R2_{= 0.44} Conditional R2_{= 0.50}

Trajectory shapes: Inhalation

Slope: 0.06 ***

Marginal R2_{= 0.31}

Conditional R2_{= 0.46}

Trajectory shapes: Exhalation

Left–Right Slope: 0.11 *** Marginal R2_{= 0.41} Conditional R2_{= 0.81} 0 1 2 3 Slope: 0.10 *** Marginal R2_{= 0.52} Conditional R2_{= 0.78} Slope: 0.06 *** Marginal R2_{= 0.33} Conditional R2_{= 0.60} Cranial–Cauda l Slope: 0.19 *** Marginal R2_{= 0.42} Conditional R2_{= 0.53} 0 5 10 15 20 0 1 2 3 Slope: 0.15 *** Marginal R2_{= 0.31} Conditional R2_{= 0.80} 0 5 10 15 20 Slope: 0.10 *** Marginal R2_{= 0.36} Conditional R2_{= 0.44} 0 5 10 15 20 Anterior–Poste rior Interfractional variability (SD) [mm]

Interfractional mean amplitude [mm]

Fig. 8.6: Interfractional mean amplitude versus interfractional variability measured by the standard deviation (SD) of

(left panel) amplitudes of the respiration-induced motion, (middle panel) trajectory shapes: marker positions relative to the trajectory centroid at the end of inhalation phase, and (right panel) trajectory shapes: marker positions relative to the trajectory centroid at the end of exhalation phases, in the left–right, cranial–caudal, and anterior–posterior directions (n = 63). The code *** means p<0.001.

tional variability of amplitudes and trajectory shapes was found to be increasing from the proximal esophagus to cardia but not always with a statistical difference between the regions (Table 8.2).

When the image modality was set as a fixed effect in the mixed-effects model, the mean differ-ence (standard error) in amplitude between the measurements in the 4D-CTs and the 4D-CBCTs was −0.1(0.1) mm (p = 0.62), −0.1(0.2) mm (p = 0.66), and 0.1(0.1) mm (p = 0.26) in the LR, CC, and AP directions, respectively. This indicates that the amplitude measured in the 4D-CT was not significantly different from that derived from the 4D-CBCT. For these 44 markers, the mean (95% CI) of the absolute differences was 0.8(0.5–1.1) mm, 1.0(0.7–1.3) mm, and 0.7(0.4–0.9) mm in the LR, CC, and AP directions, respectively (Fig. 8.7).

Statistically, the homoscedasticity and normality of the residuals of the models were satisfied for employing these linear mixed-effects models.

8.4 Discussion

This is the first study that applied retrospectively reconstructed 4D-CBCT with implanted gold markers to investigate the interfractional variability of the respiration-induced esophageal tumor

(13)

Processed on: 18-12-2018 PDF page: 128PDF page: 128PDF page: 128PDF page: 128 Proximal esophagus Middle esophagus Distal esophagus Cardia

Interfractional mean amplitude with maximum and minimum measured in the 4D-CBCT

Amplitude measured in the 4D-CT

Left–Right 0 5 10 15 Cranial–Caudal 0 5 10 15 Anterior–Posterior 0 5 10 15 6 7 8 10 11 12 13 15 16 17 18 20 21 22 23 24 Patient number Amplitude [mm ]

Fig. 8.7: Comparison between the four-dimensional (4D) computed tomography (CT) and 4D cone-beam CT (CBCT)

data in terms of the amplitudes in the three orthogonal directions for markers located in the four regions.

motion. The mean interfractional variability of amplitudes and trajectory shapes was found to be ⩽1.4 mm and the mean deviation between the amplitudes measured in the 4D-CT and in the 4D-CBCT was⩽1.0 mm. These findings suggest that the amplitudes and trajectory shapes of the respiration-induced esophageal tumor motion derived from a single acquisition of 4D-CT could be representative for that of the respiration-induced tumor motion during the treatment course.

Using markers and 4D-CBCTs to quantify the respiration-induced esophageal tumor motion has been shown to be an accurate approach. In a previous phantom study, the marker registra-tion error based on the 4D-CBCT which was reconstructed using the acquired projecregistra-tion images of 3D-CBCT was found to be <1.2 mm, comparable to that based on 4D-CT [222]. Moreover, the inter-observer variation of marker registration in the three directions based on the 4D-CBCT, which was not studied though, could be in the same order of magnitude as that based on the 4D-CT (mean:⩽0.1 mm; SD: ⩽1.0 mm) [210]. Considering the voxel size of the CTs and 4D-CBCTs and the quantified amplitudes of the respiration-induced motion, these errors can be neg-ligible but inevitable.

The linear mixed-effects model has been used in a couple of studies about motion analysis to solve the problem of non-independence in the data [224,225]. With this model, the bias of uneven number of markers and 4D-CBCT scans per patient has been properly accounted for in our study. No obvious time trend in the amplitudes allowed using the SD of the amplitudes as measure of the interfractional variability of amplitudes. The correlation between the interfractional mean

(14)

88888888

amplitude and the interfractional variability of amplitudes and trajectory shapes implies that the larger amplitude the respiration-induced tumor motion has, the more interfractional variability there could be (Fig. 8.6). Similarly, such a correlation was found in the interfractional variability of the respiration-induced motion for lung tumors [212].

The comparison of interfractional mean amplitude between the three directions and between the four esophagus regions (Table 8.2) confirms the results of the previous 4D-CT-based study [210] that the respiration-induced motion is predominantly larger in the CC direction and in the distal esophagus and cardia. Moreover, the significant difference found in the interfractional vari-ability of amplitudes and trajectory shapes between the three directions and between the four regions demonstrates that the interfractional variability is also dependent on direction and region (Table 8.2).

In spite of the direction- and region-dependence of the interfractional variability, they were generally small (mean:⩽1.4 mm). Moreover, the difference in amplitude between the 4D-CT and 4D-CBCT was limited (mean:⩽1.0 mm). These can result in a small random error (⩽1.4 mm) and therefore a limited contribution to the safety margin in all three directions and four regions [137]. Considering the 4D-CT and 4D-CBCT scan resolution and the dose grid resolution that was 2 mm in our institute, the small interfractional variability of amplitudes and trajectory shapes would not be clinically relevant. As a result, in RT of esophageal cancer, it would be sufficient to use a single acquisition of a planning 4D-CT for predicting the amplitude and trajectory shape of the daily respiration-induced tumor motion as a clinical routine.

In a couple of studies of RT for lung tumors, a stable amplitude and trajectory shape of the respiration-induced tumor motion were also demonstrated, implying that the respiration-induced lung tumor motion measured in a single 4D-CT would be representative for the motion during the treatment course [212,226]. However, for pancreatic tumors the difference in the amplitude of the respiration-induced motion between the 4D-CBCT-based and 4D-CT-based measurements was found to be >5.0 mm for 17% cases, indicating that the amplitude of respiration-induced tumor motion measured in the planning 4D-CT is not representative for that during the treatment course [203]. Considering that the pancreas is located in the abdominal region, the interfractional motion variability might be affected by a variable daily filling of the stomach, colon, and small intestines. Since the esophagus is located inside the mediastinum with more rigidness than the pancreas, it is plausible to observe a smaller difference between the amplitudes measured in the 4D-CT and in the 4D-CBCT for esophageal tumors than for pancreatic tumors.

For RT of esophageal cancer, so far bony anatomy is still the standard surrogate for setup ver-ification since an accurate tumor-based registration is not yet feasible [176]. Compared with the interfractional variability of respiration-induced tumor motion, the interfractional tumor position variation relative to bony anatomy and the intrafractional respiration-induced tumor motion were much larger and therefore would contribute the most to the planning target volume (PTV) mar-gin [176,210]. The use of internal target volume (ITV) is one way of taking into account the

(15)

respiration-induced motion of esophageal tumors in treatment planning [69,71,74]. For lung and pancreatic tumors, using the mid-ventilation approach enables the possibility of PTV reduc-tion compared to using ITV [128,202] because the moderate respiration-induced motion, as a random error, can be approximated using a Gaussian blurring [125]. Nevertheless, no dosimetric advantage of using the mid-ventilation approach in comparison with ITV has yet been investigated for RT of esophageal cancer. No matter which approach has dosimetric benefit, using 4D-CT for treatment planning is essential to account for the respiration-induced tumor motion as suggested in our previous study [210]. To monitor and correct the interfractional tumor position variation, the acquisition of daily 3D-CBCTs for online setup verification is important [176]. Furthermore, implantation of fiducial markers would be preferable for accurately quantifying the respiration-induced tumor motion and facilitating the online setup verification. Although using 4D-CBCT allows the inspection of the respiration-induced tumor motion prior to the dose delivery in each treatment fraction, it may not be necessary due to the small interfractional variability of amplitudes and trajectory shapes of the respiration-induced motion.