The irregular breathing effect on target volume and coverage for lung stereotactic body radiotherapy

Abstract The major challenge in treating a mobile target is obtaining the temporal and spatial information imaging and treatment details. This phantom study quantitatively evaluates the geometric and dosimetric effects of various treatment techniques under different respiratory patterns. The regular motion model was a sinusoidal waveform with a longitudinal range of ±1.5 cm and a period of 4 sec, while irregular motion models were generated by extracting signals from clinical cases. Helical CT for a static target and 4D CT with retrospective sorting were acquired. Phase bin, maximum, and average intensity projection (MIP and AIP) CT datasets were reconstructed. RapidArc and IMRT plans were generated on static and moving target CT datasets with different motion patterns using the phase‐based gating and nongating treatment. Dose measurements were performed using EBT3 films. Dose profile and gamma analysis (±3%/1 mm criteria) were used for dose comparisons. For the irregular motions, internal target volume variations between AIP and MIP datasets (AIP/MIP) had slight differences (−6.2% to −7.7%) for gated plans, and larger differences (−12.3% to −15.2%) for nongated plans. Dosimetric measurements showed a high gamma passing rate (>98.5%) for the static plan in the target region, while the AIP and MIP gated plans had average passing rates of 92.2% ± 5.7% and 85.8% ± 9.5%, respectively. Nongated plans had significantly lower and deviated passing rates, while the AIP and MIP plans had passing rates of 43.6% ± 22.2% and 66.7% ± 28.2%, respectively (p < 0.05). Lung stereotactic body radiotherapy treatment delivered with the gated technique did not compromise the gross tumor volumes coverage, and was insensitive to the breathing irregularities and plan techniques. Adequate margins should be accounted to cover the mis‐gating effect when using the phase‐based gating under irregular motion.


| INTRODUCTION
Stereotactic body radiotherapy (SBRT), also known as stereotactic ablative radiotherapy (SABR), is capable of delivering highly conformal radiation doses to diseases such as early-stage non-small cell lung cancer (NSCLC) and reported to provide high local control with limited toxicity. 1-3 However, the major challenge in treating a mobile target is obtaining the temporal and spatial information imaging and treatment details. Respiratory motion is patient specific, 4 and a more crucial issue is irregular breathing causes the mobile target motion pattern to vary. Unfortunately, irregular breathing is a common clinical situation. Although coaching could improve the breathing pattern reproducibility, it could not totally avoid target irregular motion during the imaging and treatment process.
The impact due to motion pattern variations could be dosimetric and geometric. Four-dimensional computed tomography (4D CT) is widely used to obtain the temporal and spatial information for a moving target. A 4D CT dataset is generally retrospective sorting with phase binning or amplitude binning. Amplitude binning is more accurate, but it is more sensitive to irregular breathing which can cause image gaps. Phase binning displays no gaps but suffers artefacts due to mis-binning. 5 To avoid missing slices from amplitude binning under different irregular breathing patterns, phase binning was used in this study. Maximum intensity projection (MIP) and average intensity projection (AIP) images created from the 4D CT phase bin datasets are usually used for treatment planning on a moving target. 6 However, irregular breathing motion in 4D CT could cause a mis-binning process and result in geometric variations, therefore significantly affecting target delineation accuracy. 5,[7][8][9][10] The main factors that affect dose delivery accuracy for a moving target are the interplay effect and the variations in patient breathing patterns during treatment. The interplay effect on an intensity-modulated dose delivery technique has been studied with proper margin and the full target motion range included, this effect is much less for clinical target volume (CTV) compared with planning target volume (PTV). [11][12][13][14][15] However, baseline shifts and irregular motion patterns have been shown to exert a remarkable influence on dose delivery accuracy. 11,16 For a realistic tumor motion treatment study, Court et al. 17   The ITV in the nongated plans included all GTV phases. In the gated plans, the ITV included only the selected GTV phases (interval of 30%-70% at the end-exhalation phases). An automatic contouring function delineated the GTVs and ITVs, which then modified manually according to a clinical procedure. Afterwards, a 5 mm uniform margin was added to the GTV or ITV to generate the PTV.
Treatment plans with the IMRT and RA techniques, and with the gated and nongated dose delivery, methods were generated. All plans were optimized with at least 95% of the PTV encompassed by the prescribed dose (D p , 6 Gy), and at least 99% of the PTV receiving doses higher than 90% of the D p . The critical organ dose-volume limits and dose conformity and gradient quality parameters were controlled according to the RTOG 0915 report. 18 The "high dose spillage" in this phantom study was much less than the criteria in this report, and had a value <1%. Including the respiratory patterns, 34 plans were generated and measured. Before each plan irradiation, a half-fan full rotation CBCT scan was performed for position verification. The AIP CT was used for image-guided in the motion target plan registration.

2.D | Geometrical and dosimetric analysis
Radiochromic EBT3 film with high spatial resolution, near-tissue equivalence and weak energy dependence were proven a viable tool for external beam dosimetry. 19,20 All films used in this study were from the same lot number. Each film sheet of 25 × 20 cm 2 was cut into smaller pieces, size of 4 × 4 cm 2 for dose-response calibrations, and 10 × 5 cm 2 for plan dose measurements [ Fig. 1(c)].
The red channel data with 16 bit digital information (pixel value, PV) were extracted and processed using the public domain software ImageJ Version 1.43 (National Institute of Health, Bethesda, MD) for dose profile comparisons, and using the FilmQA Pro software (Ashland Inc) 21 for plane dose comparisons. The net optical density (netOD) was calculated by subtracting the nonirradiation OD value: where PV Bg and PV exp are the pixel value for the unexposed (background) and exposed film piece, respectively. The sensitometric curve of EBT3 film was fitted with a third order polynomial function (netOD-to-dose polynomial function) and applied to each measurement film respectively to convert the dose.
T A B L E 1 Irregular motion models. In the static and regular motion geometrical analysis, the GTV and ITV were compared to the real target ball volume and theoretical ITV (ITV th ), respectively. The ITV th was calculated as:

Motion model
where r is the ball radius and L is the longitudinal target motion range. 23 Image artifacts and target center positions in different phases for the 10-phase and 20-phase CT images were evaluated under the regular motion condition. The target motion velocity in a sinusoidal waveform was calculated as: where p is the phase (%), where A is the amplitude, and T is the period. 24 For a regular motion pattern in this study, the maximum motion velocity was 23.56 mm/sec.
For the different irregular breathing patterns, the ITVs varied depending on the irregular breathing patterns, and calculation could not obtain the ITV th . Accordingly, the ITVs in the MIP and AIP CT images were compared to each other.

3.B.2 | Dynamic target dose delivery
The passing rates and statistical comparisons for different planning techniques, respiratory patterns, 4D CT resorting types, and dose delivery methods are listed in Table 5 The dose profile calculations and measurements for period-and amplitude-irregular patterns with gated and nongated plans in A/P and C/C axis are shown in Fig. 5. As the results showed in the 2D phases 25% and 75% also indicated that there was a beam-on imaging time delay of~0.14 sec for the CT and RPM systems used in this study. This value is similar to Smith's report. 25 The spatial accuracy of 4D CT images for a moving target seems affected mainly by the time delay effect but not the phase number. Planning CT for a moving target should avoid using only the phase bin image at the highest motion speed (e.g., phases of 25% and 75%).
The volumetric deviations for a moving target in the MIP or AIP images could lead up to 40% errors. [7][8][9]23,24 In general, the deviation becomes more severe for a smaller and faster target motion. This study evaluated the volumetric deviations for static and different motion patterns. The volume deviations were small for static and phase CT images with most deviations less than 2.5% (Table 3). Larger volume deviations (less than 7%, underestimated) were observed in the MIP and AIP images for a regular motion pattern ( Table 3).
The target volume variations between MIP and AIP images were more significant in irregular motion patterns (Table 4), and the volume underestimation was more significant in the AIP image than the MIP. For a gated plan, the volume deviation was smaller compared to the nongated plan and depended on the gating interval. For clinical applications, adequate margins should be added to the ITV in the major motion directions based on the target size and the motion irregularity when using AIP or MIP images for ITV delineation. Based on this study, 1 mm margin is adequate to cover the volume deviation for target size <4 cm dia.
To avoid dosimetric analyses motion interference, IMRT and RA plans with static CT were measured and evaluated first. The dose in the GTV was consistent between the calculations and measurements for both plans (Fig. 4), with a passing rate higher than 98.5% (Table 5). However, as described in previous studies, 26   From Table 5, these enhancements in dosimetric variations were shown in nongated plans, although there were some nongated plans with good passing rates (e.g., A-I-IMRT and P+A-I-IMRT plans with MIP CT), but statistically analyzed the nongated plans, they still exhibited significantly lower and deviated passing rates.
The interplay effect could be averaged out after several fractions. 17 However, this averaging could cause dose blurring at the field edges. 14,15 Our results demonstrated that taking a CBCT scan to reduce the positioning error, setting a proper margin to cover the full target motion range (ITV) and setup error, and using the gated dose delivery, the treatment in lung SBRT delivered by TrueBeam 6MVFFF beams did not compromise the GTV coverage (Table 5).
This finding is consistent with other studies. [11][12][13][14][15] In clinical patient treatment situations, as described in the Bo Zhao's report, 16 tumor motion may change as the baseline shift that increases with the treatment time. They concluded that a high-dose-rate mode reduced the treatment time and thus reduced the interference in the baseline shifts. The concern in a higher dose rate with more interplay effect did not emerge in this phantom study. To mitigate the baseline shift in patient treatments, a high-dose-rate mode is an appropriate choice.
Four irregular motion models were created to evaluate the dose accuracy for a moving target under irregular respiratory patterns (Table 1). With gated delivery, the AIP plan passing rates under irregular respiratory patterns were slightly lower than that of static plans.
The irregular motion patterns and plan techniques did not show a significant difference in the dose delivered to the target in the gated plans (Table 5). Treatment with phase-based gated under irregular motion pattern may trigger the beam-on and beam-off signals at the wrong phase. The mis-gating effect is managed using a margin. In this study, a 5 mm margin was added to the ITV to generate the PTV. This margin is more than 20% of the longitudinal ranges in the irregular motion models (Table 1). For gated plan, the ITV was further localized into the gated phases with relative smaller residual motion range. From the measurement results in this study, this margin is adequate to cover the mis-gating effect.
For an intensity-modulated plan, the beam intensity is modulated according to the dose constraints to the target and critical structures in the optimization process. In addition, the density distributions of these structures will affect the beam intensity map and the dose delivered. For a moving target in the AIP CT, as expected, the CT number of the ITV is "smeared" in these averaged images, and with a lower and wider density distribution along the direction of travel Their results have shown that FB and AIP plans were with similar mean effective depths but significantly different from that in MIP plans. They concluded that the AIP dataset is most favored for planning and dose calculation for lung SBRT. The dosimetric analysis in this study also showed that the AIP plans had a slightly higher passing rate than the MIP plans with the gated treatment technique (Table 5).

| CONCLUSION S
Faster target motion results in larger position errors in 4D CT image acquisition. Planning CT should avoid using only the phase bin image at the highest motion speed. The underestimation in ITV for a moving target was more significant in the AIP dataset than MIP, and is increasing with motion irregularity. This volumetric deviation is smaller for a gated plan than a nongated plan. The lung SBRT treatment delivered with the gated technique did not compromise GTV coverage and was insensitive to the breathing irregularities and plan techniques. Adequate margin should be made to cover the mis-gating effect when using the phase-based gating under irregular motion.
Nongated plans had significantly lower and deviated passing rates than the gated plans.

ACKNOWLEDG MENTS
The authors thank Ze-Jing Wang for the support of the motion platform. Research was supported by China Medical University Hospital (DMR-106-108).

CONFLI CT OF INTEREST
None of the authors has conflict of interest.