Analysis of the Amplitude Changes and Baseline Drift of Respiratory Motion during Liver Stereotactic Body Radiation Therapy Based on Intra-Fraction CBCT

Purpose: To evaluate the amplitude changes and baseline drift of respiratory motion in liver stereotactic body radiation therapy based on intra-fraction cone beam CT (CBCT). Materials and methods: Twenty-four liver SBRT patients underwent a four-dimensional computed tomography (4D CT) scan, inter-fraction cone-beam computed tomography (CBCT), and intra-fraction CBCT to evaluate the amplitude changes and baseline drift of respiratory motion. The amplitude changes were dened as the variations between the amplitude measured in the 4D CT and those measured with uoroscopy in inter- and intra-fraction CBCT. The baseline drifts were dened as the difference between the liver position errors and the setup errors in inter- and intra-fraction CBCT. Manual registration of the liver contour was performed to obtain liver position errors, and bone registration was performed to obtain setup errors. Meanwhile, the correlation among the liver position errors, the relative diaphragm position, and the amplitude changes was evaluated. Results: The systematic and random errors of the baseline drifts for intra-fraction CBCT in the medial-lateral (ML), superior-inferior (SI), and anterior-posterior (AP) views were 0.99 and 1.6 mm, 2.03 and 2.46 mm, and 1.02 and 2.07 mm, respectively. The corresponding PTV margin was 3.61 mm, 6.8 mm, and 4.00 mm, respectively. The amplitude variation ranged from -169.33% to 65.47% in inter-fraction CBCT and from -171.04% to 60.00% in intra-fraction CBCT. Inter-fraction liver position errors were signicantly related to the setup errors. The relative diaphragm position was not statistically related to the baseline drifts in the inter-fraction nor in the intra-fraction. Conclusions: Using intra-fraction CBCT, signicant amplitude variations and baseline drifts of respiratory motion were found compared with those in 4D CT. The relative diaphragm position in a single breathing cycle cannot reect the liver position during dose delivery. At least a 6.8 mm PTV margin in the SI direction is recommended to compensate for the baseline drift.

holds [4]. Gating and tracking technologies usually require additional auxiliary equipment and implantation benchmarks. Therefore, 4D CT is still the most popular method for respiratory motion management of liver SBRT with free-breathing.
After 4D CT scanning, using combination of the targets in each respiratory phases (total ten phases) to generate internal target volume (ITV) [7], using the maximum intensity projections (MIP) [8]to delineate ITV or the medium ventilation CT to delineate the target [9] are adopted in clinical practice. The 5 mm uniform PTV margins have achieved widespread use [4]. CBCT corrects the inter-fraction errors pretreatment for each fraction, and the 5 mm PTV margin is considered adequate for intra-fraction errors, which include the amplitude variations and the baseline drift of respiratory motion during treatment. At present, to analyse intra-fractional errors, post-treatment CBCT [10][11][12], KV uoroscopy [5,13], and electromagnetics [14] are used. KV uoroscopy is a form of two-dimensional imaging, analogous to electromagnetic tracking, which usually requires implantation of mental markers for veri cation. Using post-treatment CBCT in the dose delivering duration may not be a su cient surrogate for the correction of intra-fraction errors. The intra-fraction CBCT can obtain a 3D image of a patient while delivering the dose, which can represent the intra-fraction errors. Ruijiang Li, et.al. [15] used intra-fraction CBCT images and con rmed that there were large intra-fraction errors in the lung SBRT. However, until now, no study has reported intra-fraction errors for liver SBRT with free-breathing using intra-fraction CBCT.
The purpose of this study was to use inter-and intra-fraction CBCT and its two-dimensional projections to quantify the amplitude variation and baseline drift of respiratory motion during liver SBRT treatment and to evaluate whether the 5 mm PTV margin is su cient. From June 2019 to November 2019, of the 24 included patients with liver cancer, 20 were male and 4 were female, with a median age of 58 ± 11 years (range: 44-82), including 20 patients with primary liver cancer and 4 patients with liver metastases. All the patients signed a consent form before radiotherapy, and this study was approved by the Ethics Committee of West China School of Medicine, Sichuan University (No.20191128). All the patients were treated with SBRT using the VAMT technique (800 cGy × 5fractions for 11 patients, 700 cGy × 5 fractions for 5 patients, 900 cGy × 5 fractions for 6 patients, 1000 cGy × 5 fractions for 5patients). The average dose received was 3980 cGy, and the total treatment the time was 1-3 weeks.

Positioning, treatment planning, treatment
The patients were immobilized with a thermoplastic mask, which was open 5 cm below the xiphoid process and the navel to facilitate obtaining the patient's respiratory motion signal. All patients underwent a 4D CT scan with free-breathing using sixty-four slice spiral CT simulation (De nition AS, Siemens, German), with a 3 mm thickness. All patients were scanned with MRI for CT image fusion to assist in determining GTV. GTV was delineated in each phase (total ten phases) to obtain ITV, and a uniform 5 mm PTV margin [4,13] was expanded around the ITV. The average intensity projections (AIP) were adopted for treatment planning and image registration with CBCT images [13,16] (XVI Release 5.0.2, Elekta Versa HD, Stockholm, Sweden). Full arc or partial arc VMAT plans were designed using the 6 MeV (1400 MU/min) or 10 MeV (2400 MU/min) photon FFF mode.
2.3 Analysis of target (i.e., the liver) position errors The inter-fraction CBCT was acquired when the patient setup was nished; then, the manual registration between the CBCT and planning CT were performed according to the diaphragm position [17] and the liver contour [10] (See Fig. 1). Liver position errors were obtained in the three translational directions (the AP, ML and SI directions), and the rotational errors were ignored since the 6 D treatment couch was not used. Moving the couch to correct the inter-fraction errors was done if the error was greater than or equal to 2 mm in any direction. CBCT scanning was performed and registered again to verify the liver position.
If necessary (an error of more than 2 mm in any direction), the couch was moved again to correct for errors. After liver position veri cation, dose delivery was executed; at the same time, the intra-fraction CBCT scan was obtained for o ine analysis (Scanning parameters: 120 kV, 40 mA, 40 mS, 5.5frames/second; scanning start angle 181° -180° or 180° -181°).

Inter-fractions and intra-fraction liver baseline drift
To analyse the baseline drift of the liver position, all the inter-and intra-fraction CBCT images were registered with planning CT o ine. The vertebral bone was automatically registered to obtain the coarse results rst; then, the manual ne adjustment followed (See Fig. 1). The results of the vertebral body registration represented the setup errors of the patients [11]. The correlation between the setup error and the liver position errors was analysed. The differences between the setup errors and the liver position errors were de ned as the liver baseline drift. The inter-fraction and intra-fraction liver baseline drifts were de ned as: Liver baseline drift inter = Liver position inter -Setup inter Liver baseline drift intra = (Liver position intra -Setup intra ) -Liver position inter 2.5 Amplitude variation of respiratory motion and relative diaphragm position changes To quantify the amplitude variation of respiratory motion, the amplitude of respiratory motion between the 4D CT simulation and treatment fractions was compared. As for 4D CT, the amplitude of respiratory motion was measured between the end of expiration and end of inhalation using the MIM Maestro® system (Version 6.8.3, MIM Software Inc. Cleveland, USA). The inter-and intra-amplitude of respiratory motion were evaluated based on the two-dimensional projections of the inter-and intra-fraction CBCT in the anterior-posterior direction for a single breathing cycle.
To evaluate the stability of the liver (diaphragm) position, the relative diaphragm position was assessed at the end of expiration [10], and the pre-de ned vertebrae (such as T12) in the SI direction were obtained in 4D CT and the inter-and intra-fraction CBCT. The relative diaphragm position in the inter-and intrafraction CBCT was measured at the end of expiration based on the two-dimensional projections ( uoroscopy) in the anterior-posterior direction for a single breathing cycle. The correlation between the relative diaphragm position in a single breathing cycle and the corresponding liver position in the interand intra-fraction CBCT was analysed, which provided the evidence as to whether the relative diaphragm position was stable during the multi-breathing cycle (during the CBCT acquiring or dose delivering process).
To ensure the consistency of all the 3D image registrations and measurements in the two-dimensional projections, all the image registration and measurements were performed by two therapists with 10 years of experience and were reviewed by a physician with 20 years of experience in radiotherapy.

Statistical method
All data were processed and analysed using SPSS 20 and EXCEL, and correlations were analysed using Pearson correlation analysis. The PTV margin was calculated according to the formula PTV margin = 2.5 Σ + 0.7σ [18], where Σ is the systematic error, which is the standard deviation of each patient's mean; σ is the random error, which is the root mean square of each patient's standard deviation.

Result
A total of 103 paired inter-and intra-fraction CBCTs were obtained for 24 patients. The liver position errors, setup errors and baseline drifts of the inter-fraction are shown in Table 1. The systematic and random errors of liver position in the ML, AP, and SI were 1.71 and 1.85 mm, 8.93 and 5.72 mm, and 1.29 and 1.86 mm, respectively. The corresponding PTV margin of the SI direction was 26.31 mm. The maximum setup errors and liver position baseline drifts were in the SI direction, and the corresponding PTV margins were 24.37 mm and 18.50 mm. The inter-fraction liver position errors were obviously related to the setup errors. The correlation coe cients were 0.80, 0.70, and 0.625 in the ML, SI and AP directions, respectively (details see Table 2). After CBCT correction, the intra-fraction liver position errors were signi cantly reduced. The systematic and random errors in the ML, AP, and SI directions were 1.12 and 1.35 mm, 1.81and 2.91 mm, and 0.94 and 1.32 mm, respectively, with corresponding PTV margin of 3.75 mm, 6.56 mm, and 3.28 mm (Table 3). Because the couch correction is based on the result of the liver position errors, there were still large interfraction setup errors. The maximum error was in the SI direction, with a systematic error of 5.87 mm and random error of 3.83 mm. However, the correlation between the intra-fraction liver position errors and intra-fraction setup errors was low. Only in the ML and SI directions was there a statistical correlation; the correlation coe cients were 0.46 and 0.25 (for details, see Table 2). The systematic error and random error of the intra-fraction liver position baseline drift in the SI direction were 2.03 and 2.46 mm, respectively, and the corresponding PTV margin was 6.8 mm (see Table 3). The average respiratory motion in 4D CT, the inter-fraction CBCT and the intra-fraction CBCT were 13.39 mm, 13.12 mm and 12.09 mm, respectively. However, the amplitude variation of respiratory motion in the same patient was considerable. Taking the amplitude of the respiratory motion in 4D CT as a reference, the amplitude variation ranged from-169.33-65.47% in the inter-fraction CBCT and from − 171.04-60.00% in the intra-fraction CBCT (see Fig. 1). The difference of the respiratory motion amplitude in 4D CT and in the inter-fraction CBCT (103 fractions) was 0.28 ± 5.41 mm; 66 of 103 fractions (64%) were within 5 mm. The difference of the respiratory motion amplitude in 4D CT and in the intra-fraction CBCT (103 fractions) was 1.03 ± 4.35 mm, with 69% of fractions being within 5 mm (See Fig. 2). The relative diaphragm position (measured by the two-dimensional projections at the anterior-posterior direction for a single breathing cycle) was not statistically related to the baseline drift in the inter-fraction nor the intra-fraction (See Table 4). The respiratory motion amplitude had a weak statistical correlation to the liver position error in the SI direction both in the inter-fraction and the intra-fraction (See supplementary le 2).

Discussion
Chan M reported [9] that there was no signi cant difference between the use of 3 D CBCT and 4 D CBCT in liver radiotherapy. To reduce the CBCT scanning time, we used 3 D CBCT image guidance. Markus Oechsner reported [16]that AIP is more suitable for 3D CBCT than MIP. Bedos L et al. also believed that [13] CBCT was an average image and that registration with the planned AIP image can represent the patient's breathing pattern, so we used AIP as the reference image for registration with 3D CBCT images for liver SBRT. In the inter-and intra-fraction CBCT, the liver contour can be clearly displayed (Fig. 1), and the registration error of the two therapists (inter-observer error) was less than 1 mm in the ML, SI and AP directions (SD).
Our results showed that the liver position and the setup errors for the inter-fraction were considerable, and there was a signi cant correlation between the setup errors and the liver position errors. Meanwhile, the liver baseline drifts demonstrated large systematic and random errors in the ML, SI and AP directions, which were 1.25 mm and 1.64 mm, 6.12 mm and 4.57 mm, and 1.41 mm and 1.79 mm, respectively. Case RB et al. [10] found that in patients with free breathing, the systematic error and random error of the liver position baseline drifts in the ML, SI and AP directions were 1.5 mm/1.8 mm, 3.1/3.6 mm, and 1.6/2.7 mm, respectively. The error in the SI direction in that report was signi cantly smaller than ours. The main reason for this difference was that in that report, the patients planning used end-expiratory CT and the same respiratory phase CBCT image for registration. In our study, the free breathing average CBCT and AIP CT were used for registration. Dhont J reported [19] that the liver baseline drifts between inter-fractions were 1.6 ± 1.3, 3.0 ± 1.2 mm, and 2.1 ± 1.4 mm in the ML, SI, and AP directions, respectively.
In our study, the liver position errors and baseline drifts in the intra-fraction were signi cantly reduced compared to those in the inter-fraction, indicating that using CBCT to correct the liver position was effective. The baseline drifts in the ML, SI, and AP directions were 0.99/1.60 mm, 2.03/2.46 mm, 1.02/2.07 mm, respectively, similar to the Case RB report using CBCT (1.2/2.2 mm 1.4/3.0 mm 1.0/1.9 mm, respectively) [10]. Bedos L. et. al. [13]used a KV plain lm to acquire the end-expiratory image to analyse the patient's intra-fraction errors and found that the 99% intra-fraction error in the SI direction was within 5 mm, which was smaller than in our study. However, the images were collected before the treatment beam was on, and the acquisition time was limited, which cannot represent intrafraction errors. Another report showed that the medians (ranges) of the baseline drifts were 1.87 (0.06-12.04) mm, 0.35 (0-3.39)mm and 1 (0.02-7.21)mm in the SI, LR and AP directions, respectively [20]. However, this study de ned the baseline as the average position in the breathing cycle during treatment. In our study, the baseline was de ned in 4D CT, and the intra-fraction CBCT was used to evaluate intrafraction errors. The average frame number of the intra-fraction CBCT was 467 frames, and the average acquisition time was 84.9 seconds, which can include approximately 20 breathing cycles. Xu Q et al. [21]used liver markers to evaluate the baseline displacement, and they found that the baseline displacement of ML, AP and SI were 2.1 ± 2.3 mm, 2.9 ± 2.8 mm, and 6.4 ± 5.5 mm, respectively. By using 4D CT as a reference, the baseline drifts were signi cant during treatment according to these reports and our ndings. In other words, the vertebral bone cannot be used for position veri cation for liver SBRT due to the considerable liver baseline drifts.
CBCT using two-dimensional projections ( uoroscopy) can evaluate the stability of respiratory motion during the entire respiratory cycle. Dhont J et al. [19] used a single metal marker to analyse the SI direction of respiratory motion The amplitude change over 5 mm was 53% for the inter-fraction and 28% for the intra-fraction. Shimohigashi Y S et al. [22] reported that in using abdominal pressure, the average movement in the SI direction was 5.3 mm, and the maximum value was 14.8 mm. The medians (ranges) of the intra-fraction amplitude variation across all patients were 4.3 (1.6-6.0) mm, 0.5 (0.2-2.2) mm and 1.5 (0.3-3.3) mm in the SI, LR and AP directions, respectively [20]. In the current study, the variation of the respiratory motion amplitude among 4D CT, the inter-fraction CBCT and the intra-fraction CBCT in the SI direction was small (the average amplitude was 13.39 mm, 13.12 mm and 12.09 mm, respectively), indicating that the overall patient respiratory motion amplitude was relatively stable. However, for individuals, there was a large difference. The variation of the respiratory motion amplitude for the intrafraction CBCT ranged from − 171.04-60.00%, and the overall difference was 1.03 ± 4.35 mm. In 69% of cases, the overall difference was less than 5 mm. However, because we only evaluated a single breathing cycle using uoroscopy, this cannot represent the real respiratory motion state during the whole treatment process (it was a snapshot image like 4D CT). The respiratory motion amplitude and baseline may change during the treatment process, and this uncertainty was more obvious when the movement was greater than 7 mm [19]. Meanwhile, the report noted that the accuracy of the 4D CT phase classi cation according to amplitude was higher than that of the phase classi cation according to time [23]. Therefore, if 4D CT was used, the change in the motion amplitude and the baseline drifts should be completely evaluated and understood when determining the PTV margin.
Considering the baseline drifts, Dhont J et al. recommended the 8 mm PTV margin for liver [19].Worm ES et al. [24]reported that when the target area was delineated using 4D CT with moderate ventilation, a 10 mm PTV margin in the SI direction and 5 mm AP and ML directions should be used. Case RB [10] found that the patient's respiratory motion changed little during radiotherapy, and the results were not statistically related to the time of radiotherapy. Although there were large variations of the respiratory motion amplitude in our study, this cannot be a main reason to expand the PTV margin. Fortunately, the respiratory movement is a 3D movement around the baseline mainly in the SI direction. The baseline drift was a systematic error, which can induce a systematic dose deviation, and the amplitude changes were random errors, which can blur the dose around the target edge. Therefore, the baseline drift in the intrafraction is the most important error when determining the PTV margin. According to the results in our study, the 5 mm PTV margin was su cient in the ML and AP directions, but there needs to be a 6.80 mm PTV margin in the SI direction.
Since metal markers were not implanted in or near the tumour, the three-dimensional position deviation of the tumour cannot be analysed using two-dimensional projections. Therefore, implanted metal markers [25]or MR [3]should be used to obtain more soft tissue information and to further analyse the respiratory motion amplitude changes and baseline drifts. If we want to comprehensively analyse the variation of respiratory motion, we need automatic methods to evaluate the two-dimensional projections during whole treatments.

Conclusions
When using intra-fraction CBCT, signi cant amplitude variations and baseline drifts of respiratory motion were found compared with those in 4D CT. The relative diaphragm position in a single breathing cycle cannot re ect the liver position during dose delivery. At least a 6.8 mm PTV margin in the SI direction is recommended to compensate for the baseline drift. The dose effect caused by the respiratory motion amplitude changes and baseline drifts during dose delivery should be studied further.   Respiratory amplitude measured using 4D CT, the inter-fraction CBCT and the intra-fraction CBCT.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.