Dosimetric impact of tumor position displacements between photon and proton stereotactic body radiation therapy for lung cancer

Abstract

Purpose

To investigate the impact of tumor position displacements (TPDs) on tumor dose coverage in photon and proton stereotactic body radiation therapy (SBRT) treatments for lung cancer patients.

Methods

From our institutional database of 2877 fractions from 770 lung cancer patients treated with photon SBRT in 2017-2021, 163 fractions from 88 patients with recorded iso-center shifts of >1.5 cm in any direction under kV-cone-beam CT guidance were identified. By double registrations with bony and tumor alignments, the difference between the iso-center shifts of these two alignments was categorized as TPDs. One fraction from each of 15 patients who had TPD magnitudes >3 mm were selected for this study. For each patient, one proton plan using intensity modulated proton therapy (IMPT) with robust optimization was generated retrospectively. All photon plans had V_100%RX>99% of GTVs and V_100%RX>98% of ITVs. Proton plans were evaluated with two worse-case scenario (voxelwise worst and worst scenario) using 5mm and 3.5% uncertainty to achieve the same planning goals as the corresponding photon plans. These two evaluation proton plans were named proton-1st and proton-2nd plans. The dosimetric effect of TPD was simulated by shifting tumor contours with the corresponding shift on patient specific planning CT and by recalculating the dose of the original plan.

Results

The range of magnitude of TPDs was 3.58–28.71 mm. In photon plans, TPDs did not impact tumor dose coverage, still achieving V_100%RX of the GTV≥99% and V_100%RX of the ITV≥98%. In proton plans for patients with TPDs>10 mm, inadequate target dose coverage was observed. More specifically, 8 fractions of proton-1st plans and 4 fractions of proton-2nd had V_100%RX of the GTV<99% and V_100%RX of the ITV<98%.

Conclusions

Adequate tumor dose coverage was achieved in photon SBRT for magnitude of TPDs up to 20 mm. TPDs had greater impact in proton SBRT and adaptive planning was needed when the magnitude of TPDs>10 mm to provide adequate tumor dose coverage.

INTRODUCTION

Stereotactic body radiation therapy (SBRT) involves precise delivery of high doses of radiation to discrete targets in five or fewer fractions while minimizing dose to healthy tissues.[1] SBRT, providing increased local control and survival rate, compared to conventional radiation, is commonly used to treat early stage non-small cell lung cancer (NSCLC) patients.[2-4] Conventionally, SBRT is delivered using photon beams. With the advent of proton pencil beam scanning (PBS) intensity-modulated proton therapy (IMPT), more conformal dose distribution can be achieved in proton SBRT.[5,6] While the unique beam characteristics in proton therapy offer further normal tissue sparing, such a technique may be greatly impacted by the interplay effect and range uncertainties of particle beams, leading to underdosing of target or overdosing of healthy tissues.[7,8] Historically, a major concern in thoracic radiotherapy delivery has been tumor motion due to breathing and variation in lung anatomy due to medical comorbidities.[9,10] Currently, respiratory motion can now be taken into account by a range of tumor motion management strategies, such as four-dimensional computed tomography (4DCT), breath-holding, abdominal compression and gating.[11] Anatomical changes in the lung, such as density changes and tumor position displacements (TPDs) from atelectasis, pleural effusion and pneumonia/pneumonitis are difficult to predict and are not typically considered in the planning process.[12] Such changes may however influence dosimetric results and adaptive radiotherapy (ART) is often used to alleviate dose variations due to anatomical changes.[13,14] Kilo-voltage cone beam CT (kV-CBCT) based image-guided RT (IGRT) is employed in SBRT prior to the treatment to account for daily changes in patient positioning and target anatomy. Because of this daily assessment, CBCTs may be used to assess the potential need of ART.[15] Bertholet et al. [16] reported survey results looking at the ART practice in 177 centers worldwide and found that lung was the second most common site requiring ART (36% of respondents). When it comes to comparing photon and proton beams, due to the sensitivity of proton dose deposition in the presence of uncertainties along the beam path, ART is used more frequently in proton therapy than intensity-modulated radiation therapy (IMRT).[17,18] To understand the rate of needing ART, Chen et al. concluded a retrospective cohort study of tumor anatomical change in NSCLC patients treated with IMRT or proton therapy.[18] They found that plan updates may be required when the 3D shift of the mean clinical target volume (CTV) from the planning CT (pCT) exceeded 8.26 mm. In the previous study by Chen et al. which focused on late staged NSCLC pateints treated with conventional treatment fractions, tumor volumes were large and changed during radiotherapy. Different from their study, this present work is to examine patients with early stage NSCLC receiving hypo-fractionated SBRT and therefore no change in tumor volume is assumed.[19] Patient setup uncertainties and tumor intra-fraction motion are beyond the scope of this study because TPDs are the result of anatomical changes intrinsically. Furthermore, robust optimization for IMPT does not account for anatomical changes during optimization [20] and no established guideline in proton robust evaluation is currently available.[21] This retrospective study is to quantitatively investigate dose variations of targets due to TPDs based on soft-tissue matching considering photon and proton SBRT. We also inspected dosimetric impact of proton plans with two different criteria in robustness evaluation.

METHODS AND MATERIALS

Patient selection

From our institutional database of 2877 fractions from 770 lung cancer patients treated with photon SBRT for the interval 1/1/2017-4/19/2021, 163 fractions from 88 patients with daily iso-center shifts of >1.5 cm in any direction according to kV-cone-beam CT (kV-CBCT) guidance were selected. TPDs of these fractions were identified according to the difference between the iso-center shifts of bony and tumor alignments (addressed in the following subsection). One fraction from each of 15 patients (2%) who had magnitudes of TPDs >3 mm were chosen for this study. Only patient 15 required adaptive radiotherapy. Doses employed are listed in Table 1. Doses in proton therapy are prescribed as Gy (RBE).

Table 1.

Patient, ITV volume, prescription and tumor position displacement characteristics for 15 patients.

Patient #	ITV (cc)	Prescription (Gy)	ML (mm)	AP (mm)	SI (mm)	TPD (mm)
1	36.37	50	1.15	0.11	3.39	3.58
2	12.48	50	1.73	2.25	2.24	3.62
3	1.87	48	0.24	3.86	4.81	6.17
4	18.52	60	1.68	3.59	4.96	6.35
5	13.56	54	2.15	0.15	6.34	6.70
6†	2.34	60	4.01	0.92	8.47	9.42
7†	3.42	54	0.64	9.86	1.17	9.95
8	13.97	34	2.08	7.69	6.31	10.16
9	15.92	48	4.45	9.42	0.9	10.46
10	13.85	54	0.77	10.2	4.07	11.01
11†	15.39	60	2.39	5.34	10.57	12.08
12	5.08	48	0.67	14.74	1.49	14.83
13	63.16	54	7.57	10.87	8.06	15.51
14	4.68	34	8.13	15.06	11.62	20.69
15	15.06	54	6.48	23.91	14.52	28.71

†ABC

Imaging acquisition

This study population consisted of 12 patients who underwent 4-dimensional (4D) CT scanning and 3 patients who underwent CT scanning with Active Breathing Coordinator^TM (ABC). According to our institutional guideline of motion managements for SBRT, the magnitude of the breathing induced tumor motion observed from 4DCT acquired at simulation should be within 1.0 cm and the magnitude of breath-hold reproducibility observed from three repeat ABC scans acquired at the simulation should be < 0.5 cm. If tumor motion exceeds the tolerance, the physicist should inform the physician and discuss an alternate motion management option such as gated treatment or breath-hold treatment. Average intensity projection (AIP) datasets were generated from 10 respiration phases of 4DCT images, while one representative of 3 acquired scans was used for ABC. The internal target volume (ITV) was contoured using the full range of motion on the 10 respiratory phased or 3 separate ABC scans to encompass the range of motion of the tumor. Planning target volume (PTV) was contoured by expanding ITV with an isotropic margin of 5 mm. Before each treatment, kV-cone-beam (kV-CBCT) images were acquired for all patients and a soft tissue alignment was used to localize the tumor and correct for setup errors. These were reviewed and adjusted by both a physicist and physician in cooperation prior to treatment delivery. The final 6D couch shifts were recorded in the Record-and-Verify (R&V) System (MOSAIQ v2.64, Elekta, Sunnyvale, CA, U.S.A.). If the couch shifts were greater than 1.5 cm in any directions, a second kV-CBCT was required for verification after applying couch shifts.

Imaging registration and tumor position displacement

For the purpose of this study, we retrospectively performed the rigid image registration between pCT and kV-CBCT using MIM software (MIM Maestro v6.9.7, Cleveland, OH) to calculate TPDs. We first aligned the kV-CBCT to the pCT using box-based alignment at the tumor volume level, which mimicked the bony alignment. The fused images were visually inspected and manually adjusted if the automatic registration did not give satisfactory results. An example of image fusion is shown in Figure 1. The original ITV contour associated with pCT is outlined in purple. In the overlay mode, the fusion setting slider was used to set the transparency of the pCT and kV-CBCT, where the upper and lower panels of Figure 1 represented the pCT and kV-CBCT, respectively. We copied the original tumor contours (GTV and ITV) as well as the isocenter associated with the pCT and aligned them to the tumor location on the kV-CBCT, named shifted contours (outlined in blue in Figure 1). For simplicity, Figure 1 only displays ITV contours. TPD was defined as the difference between the isocenter positions of the original and shifted tumor contours. , where ∆X, ∆Y and ∆Z were the translation shifts in RL, AP and SI directions of the isocenter position, respectively. Note that patient setup positioning was verified by both physicist and physician prior to the treatment delivery, as described in the previous subsection. TPD is patient specific, mostly due to atelectasis, pleural effusion and pneumonia/pneumonitis. TPD only occurs to 2% of our SBRT lung patients and cannot be predicted from 4D-CT, repeat ABC scans.

Figure 1.

Representative fusion images between pCT (upper) and CBCT (lower) after bony alignment of patient #7. The original ITV contour associated with pCT is outlined in purple and the shifted ITV is outlined in blue. Spinal cord contour is shown in green

Treatment planning

Photon plans

All clinical plans were planned with the Pinnacle³ treatment planning system (TPS) (Philips Medical System, Cleveland, OH) for Varian Edge machines (Varian Medical Systems, Palo Alto, CA). VMAT plans were generated with 6 MV FFF beams and 120HD MLC. These plans were optimized to ensure that >99% of the GTV, >98% of the ITV, and >95% of the PTV received the prescription dose. For this study, we imported all clinical plans into Raystation (RaySearch Laboratories, Stockholm, Sweden) using DICOM radiotherapy (RT) import for recalculation. CT density table used in Pinnacle was imported to Raystation and a generic Varian Edge machine was used for plan calculations. Final dose calculation was performed using Collapsed Cone algorithm and a dose grid size of 3 mm. Due to different TPSs reconstructing dose distributions and structures differently, clinical RT dose was imported along with each RT plan. After recalculating the RT plan with the same monitor units (MUs) from the clinical plan, the recalculated photon plan was normalized to have the same percentage of the PTV receiving the prescription dose (V_100%RX) as that in the RT dose. To provide the most clinical relevant evaluation, we chose to re-calculate as opposed to re-optimize the clinical photon plans in Raystation.

Proton plans

All proton plans were retrospectively generated in Raystation using IMPT technique with multi-field optimization (MFO). A generic cyclotron-based scanning proton machine with energies ranging from 70-230 MeV was used. ITV-based robust optimization with 5 mm setup uncertainty in RL, AP and SI directions combining a density uncertainty of 3.5% was conducted in all plans. For patients who underwent 4DCT scanning, 4D-averaged CTs were used for 3D robust optimization. Final dose calculation was performed using Monte Carlo algorithm with 1% uncertainty and a dose grid size of 3 mm. All plans were optimized using two or three beams per target. Beam angles were chosen to best cover the target while sparing normal lung tissues and critical structures. Anterior and lateral beams were avoided if possible in order to minimize the change in anatomy along the beam path. In the nominal scenario, all proton plans had V_100%RX of the GTV>99% and V_100%RX of the ITV>98%. In the voxelwise worse scenario, the nominal proton plans were normalized to have V_95%RX of the ITV>98% and these plans are named proton_1st. In the worst scenario, the nominal proton plans were normalized to have V_100%RX of the ITV>98% and these plans are named proton_2nd. For proton plan evaluations, 12 scenarios were evaluated including combined six setup shifts along the axes (5 mm) and two range uncertainties (3.5%). Not every proton plan had adequate target dose coverage after optimization, and therefore re-normalization was needed. All of our proton plans met the machine MU limitation.

Plan verification of tumor position displacement

In this paper, ‘original plans’ refer to plans calculated on the pCT with original tumor contours. To evaluate the dosimetric impact of TPDs, we copied the original plan’s beam data to the pCTs with shifted tumor contours. HU override was applied to the shifted ITV (0.7 g/cm³) and to the original ITV minus the overlap with the shifted one (0.26 g/cm³). We assigned the radiation beam isocenter to the isocenter of the shifted tumor and performed dose calculation without further optimization. ‘Verification plans’ refer to plans calculated on the pCT with shifted tumor contours.

Plan evaluation

To compare the plan quality of original plans among the photon recalculation and two proton plans, conformity index (CI), RTOG CI, volume receiving 50% of the prescription dose (V50%), the percentage of normal lung tissues receiving 20 Gy (Lung V20Gy) and the mean normal lung dose were accessed. CI is defined as TV_RX/V_RX, where TV_RX is the volume of ITV receiving the prescription dose and V_RX is the tissue volume receiving the prescription dose. RTOG CI is defined as V_RX/V_TV, where V_TV is the ITV volume. The paired Wilcoxon signed rank test was used to compare the plan quality among photon recalculation, proton_1st and proton_2nd plans, with p<0.05 considered as statistically significant. To examine the verification plans, V_100%RX of the GTV>99% and V_100%RX of the ITV>98% were used for target dose coverage evaluation.

RESULTS

Photon plan recalculation: clinical vs. recalculated plans

Figure 2 shows MU relative difference between clinical photon plans from Pinnacle and recalculated photon plans from RayStation, defined as 100% * (MU (Pinnacle) – MU (RayStation))/MU (Pinnacle). The average of absolute MU relative difference was 4.08%. For Patient 14, a large MU relative difference of 9.89% was observed. Such large difference could be attributed to small tumor volume (ITV=4.68 cc; PTV=18.67 cc) and higher plan modulations with high total MU of 8496. The use of a generic machine from RayStation could also contribute to MU discrepancies in recalculation.

Figure 2.

MU relative percentage difference between clinical photon and recalculated photon plans

Plan quality comparisons of original plans

Dosimetric endpoints of original plans for recalculated photon and two proton plans are plotted in Figure 3, including (a) CI, (b) RTOG CI, (c) V50% (cc), (d) Lung V20Gy (%) and (e) Mean lung dose (cGy). For target conformity, proton_1st plans produced the highest degree of conformity compared to photon recalculation and proton_2nd plans with significant differences (p<0.0001) in both CI and RTOG CI. Plan dose falloffs considering V50% was not significantly different between recalculated photon and two proton plans. The median V50% for photon recalculation, proton_1st and proton_2nd were 161.02, 121.97, 125.90 cc, respectively. No statistically significant difference was found in lung V20Gy between photon recalculation and proton plans, while both proton plans showed a significantly reduced mean lung dose compared with photon recalculation plans (p<0.0001). The median Lung V20Gy/mean lung dose were 3.55%/312 cc, 3.21%/191 cc, and 3.4%/203 cc for recalculated photon, proton_1st and proton_2nd, respectively. All the dosimetric endpoints evaluated showed statistically significant differences (p<0.0001) between two proton plans.

Figure 3.

Dosimetric endpoints, including: a) conformity index (CI); b) RTOG CI; c) V50% (cc); d) LungV20Gy (%); e) Mean lung dose (cGy) of recalculated photon and two protons plans calculated with the original ITVs

Dosimetric effect of tumor position displacement

The mean magnitude of spinal cord displacements was 0.63 mm (range: 0.34-1.18 mm) after performing the rigid registration in MIM, which demonstrated that set-up errors based on bone match were minimal. The range of magnitude of TPDs in this study was 3.58-28.71 mm. The magnitudes of TPDs in ML, AP and SI directions for all patients were listed in Table 1. The median magnitudes were 3.10 mm in ML, 7.98 mm in AP and 5.57 mm in SI. Large variation in tumor positional change was observed especially in the AP direction, where 9 fractions had displacements >5 mm. Figure 4 shows target dose coverage of verification plans with TPDs for recalculated photon, proton_1st and proton_2nd for all patients. For proton_1st, target with V_100%RX of the GTV <99% and V_100%RX of the ITV <98% was seen in eight fractions as shown in grey diagonal pattern. For proton_2nd, four fractions had target under-dosage, as shown in yellow diagonal pattern. In proton SBRT, decreases in target dose coverage can be seen when the magnitude of TPDs >10 mm. For those eight fractions that had inadequate target dose coverage in proton_1st, the V_100%RX of the GTV and ITV were increased on average by 9.41% and 8.12%, respectively, in proton_2nd. On the other hand, all fractions of verification plans for recalculated photon had V_100%RX of the GTV ≥99% and V_100%RX of the ITV ≥98% for magnitude of TPDs up to 28.71 mm (patient #15), which was the largest shift in our study. Note that patient 15 had adaptive radiotherapy although the target dose coverage was maintained in the recalculated photon plan. Figure 5 shows an example of dose distributions (patient #8) between the original and verification plans for recalculated photon and two proton plans. The original and shifted ITV associated with the pCT were shaded in purple and blue, respectively, as seen in Figure 5. In original plans, V_100%RX of the ITV were 100% for recalculated photon, 99.24% for proton_1st and 100% proton_2nd. In verification plans, both recalculated photon and proton_2nd plans had V_100%RX of the ITV ≥98%, while V_100%RX of the ITV was 94.22% for proton_1st. To compare the original and verification plans between patients with the smallest and largest TPDs in this study, we plotted the dose volume histograms (DVHs) of the ITV for photon, proton_1st and proton_2nd plans in Figure 6. Patient 1 with the smallest TPD is shown in the upper panel while Patient 15 with the largest is shown in the lower one. A larger variation between the original plans (solid line) and verification plans (dashed line) is observed for the case with larger TPD.

Figure 4.

V_100%RX of GTV (left) and ITV (right) of verification plans for photon recalculation, proton_1st and proton_2nd. The corresponding magnitude of tumor position displacement is shown in scatter points associated with the y-axis on the right side. Fractions had V_100%RX of the GTV <99% or V_100%RX of the ITV <98% were shown in diagonal patterns

Figure 5.

Dose distribution comparisons (Patient 8) between the original and verification plans with tumor position displacements for photon recalculation, proton_1st and proton_2nd. Original and shifted ITV were shaded in purple and blue, respectively. Prescription was 34 Gy

Figure 6.

Dose volume histograms (DVHs) of the target (ITV) between the original (solid line) and verification (dash line) plans for photon recalculation (black), proton_1st (blue) and proton_2nd (red). The upper and lower panels show plans with the smallest (Patient 1) and largest (Patient 15) tumor position displacements, respectively, in this study

DISCUSSIONS

This retrospective study investigates dose variations between photon and proton SBRT for lung cancer patients with TPDs by comparing dose distributions. When the magnitude of TPDs exceeded 10 mm, the plan verification process should be performed for proton plans to ensure adequate target dose coverage. On the other hand, photon plans were more robust against tumor positional change, where sufficient target coverage could be achieved without re-planning for magnitudes of TPDs of up to 20 mm. Proton plans evaluated with different criteria in robustness evaluation produced different results in the plan verification process to account for TPDs. By incorporating robustness evaluation in proton plans, dose variations due to tumor positional shift can be alleviated.

Compare to respiratory induced tumor motion, anatomical change was reported to have a greater impact on target dose coverage in IMRT for patients with NSCLC.[10] Although breathing interplay effect should be carefully considered during lung proton therapy,[8] it is beyond the scope of the present study. This study focuses on TPDs, which are resulted from changes in patient anatomy. Previous studies of photon therapy showed that the primary TPDs relative to bony anatomy could be >10 mm due to atelectasis,[14] and implementing ART for NSCLC might have dosimetric and clinical benefits.[10,12-14,22] In the study reported by Chen et al.,[18] the time-averaged mean CTV variations from the pCT for patients who were in the non-ART/ART group were 5.19±1.59/9.24±1.85 mm (p<0.001) using proton therapy and 6.61±2.82/8.63±2.57 mm (p=0.156) using IMRT. They found that IMRT was more robust to anatomical changes compared to proton therapy, which was consistent with our finding in this study. Imaging data from photon therapy was used for the retrospective proton study in the present work, which was confirmed by Chen et al., [18] where no difference in imaging responses of targets was found for patients treated with the two different modalities.

Li et al. [20] showed that compared with PTV-based planning, robust optimization in IMPT could reduce the dose variation due to anatomical changes but may not be adequate to account for changes during treatments. They also reported that the optimization technique used during planning would be less relevant if an ideal robustness evaluation tool is used and the level of acceptable robustness is defined. Currently, there is no consensus in proton robust evaluation. The acceptable robustness metrics can be dependent on tumor locations and volumes as well as physician/physicist reviews. In the robust evaluation module in RayStation, dose distributions can be assessed using worst scenario and voxelwise worst scenario. The former considers a single scenario, while the latter takes all the scenarios into account.[21] In our clinical practice, V_100%RX of the GTV≥99%, V_100%RX of the ITV≥98% and V_100%RX of the PTV≥95% are the metrics used to assess target dose coverage for photon SBRT. V_95%RX of the CTV≥99% and V_95%RX of the PTV≥95% are commonly used in photon and proton plan evaluation for lung cancer.[8,23] These two coverage criteria were therefore chosen for dose distribution evaluations in this study. Here, we found that these two proton plans produced distinct results in verification plans with TPDs although both had adequate target dose coverage in original plans. Yet, robust optimization technique is not developed to account for anatomical changes, and by incorporating robust evaluation, improved target dose coverage due to TPDs can be achieved. Nevertheless, the plan verification process described in our study is still highly recommended for patients with TPD magnitudes of >10 mm undergoing proton SBRT. One fraction from patient #2 had shown target underdosage even when the magnitude of TPD was only 3.62 mm. One possibility is that the generic proton machine used in this study had a spot size in air at the isocenter of 2.73-7.00 mm, corresponding to energy of 230-70 MeV. When a proton machine with a smaller spot size is used, more uniform dose may be achieved for targets having smaller volumes. Another reason could be the proton planning technique. A more robust plan for smaller target may be achieved using single-field optimization (SFO), which is beyond the scope of this present study and can be pursued in future research. In our analysis on the comparisons of original plans, all the dosimetric endpoints showed a statistically significant difference between the two proton plans. Compared with photon SBRT, proton_1st plans generated better conformity and similar dose fall-off and dose to normal lung tissue, while proton_2nd plans had comparable plan quality in general. When comparing the two treatment modalities in SBRT for lung cancer, photon plans were more robust to TPDs and proton could generate comparable or better plan quality compared to photon therapy.

Dosimetric calculation based on CBCT images in radiation therapy can be done in numerous ways, including pCT/CBCT calibration, image registration, HU override and dose deformation, etc.[24] For CBCT dose recalculation in proton therapy, deforming the pCT onto CBCT via deformable image registration (DIR) was used in most studies.[24] Veiga et al. studied lung cancer patients using a proton gantry-mounted CBCT system and found that images generated from deforming the pCT onto the daily CBCT followed by a correction could be an accurate alternative to dose recalculated on repeat 4D CT.[17] They suggested that accurate HU information instead of accurate voxel-by-voxel mapping demonstrated a greater impact on dose calculations. However, DIR may not provide promising results when there is large change in anatomy giving rise to discrepancies between CBCT and deformed image.[25] The shifted tumor contour could be alternatively done by manually contouring targets on daily CBCT images. In this case, re-planning would be required due to target volume variations, which was not appropriate for the current study design. In this study, we evaluated verification plans by manually shifting tumor contours on the pCT due to the sensitivity of proton dose computation towards density resolution. Such method may have its limitations which we observed when the tumor was close to organs at risk (OARs). Due to the tumor location being close to the heart, the shifted ITV contour was superimposed over heart tissue on the pCT although it did not overlap the heart contour that was shown in orange in the upper panel of Figure 7. The density variation on the CT images at the interface of heart and the shifted ITV contour might have introduced uncertainties to the dose calculation. To further examine this particular case, we performed another dose calculation based on CBCT images which contained patient’s anatomy on the day of treatment. An additional bony structure was created using semi-automated density-based contouring in MIM. We then aligned CBCT to pCT using rigid image registration and transferred all the contours including the shifted ITV to the CBCT. We manually corrected the lung contour to match the anatomy in CBCT. Beams from the original plan were copied to the CBCT image set and HU override was applied to the body contour (1.0 g/cm³), shifted ITVs (0.7 g/cm³), lungs (0.26 g/cm³) and bony structure (1.85 g/cm³) for dose calculation. The dose distribution comparison of the verification plans calculated on pCT and CBCT was shown in Figure 7. V_100%RX of the ITV calculated on pCT/CBCT were 99.02%/99.46% for recalculated photon, 92.20%/90.14% for proton_1st and 97.42%/96.15% for proton_2nd. The average heart dose for pCT/CBCT were 409 cGy/424 cGy for recalculated photon, 50 cGy/75 cGy for proton_1st and 50 cGy/74 cGy for proton_2nd. In the original plans, V_100%RX of the ITV/averaged heart doses were 98%/307 cGy, 100%/14 cGy, and 100%/14 cGy for recalculated photon, proton_1st, and proton_2nd, respectively. This result suggested that dose calculation with TPD on CBCT had larger variation compared to that on pCT for proton plans. Anatomical changes during radiation therapy may induce variations in radiation doses to lung and heart [26] and affect target dose coverage. Plan verification is highly recommended when targets are adjacent to OARs. For some patients, TPDs were observed in more than one fraction over the course of treatment. However, we only analyzed one fraction due to similar results from the same patient’s anatomy. This study is limited by small sample size and lack of evaluation of OAR dose due to the majority of tumors being away from critical structures. The use of pCT in the present study provides high image quality but may not capture a patient’s daily anatomy. A hybrid photon-proton radiation proposed by a previous study [27] could be a promising approach when considering anatomical changes and delivery uncertainties, where sharper dose falloff can be achieved in proton therapy and photon provides better robustness. High precision and accuracy in treatment delivery for SBRT is essential to account for daily changes in target anatomy. Future studies investigating dose variation using different CBCT dose calculation techniques or on repeated CT are needed, particularly for proton therapy.

Figure 7.

Verification plans of patient 13 with tumor position displacement on pCT vs. CBCT for photon recalculation, proton_1st and proton_2nd. The shifted ITV is shaded in blue. Heart volume is contoured in orange. Prescription was 54 Gy

CONCLUSIONS

In proton SBRT for lung cancer patients, adaptive therapy may be needed when magnitude of TPDs 10 mm to provide adequate target dose coverage. Incorporating proton robust evaluation may improve target dose coverage, and guidelines for proton robust evaluation are needed. Photon SBRT was more robust against tumor positional changes, where adequate target dose coverage was achieved for magnitude of TPDs up to 20 mm.