Feasibility of Pencil Beam Scanned Intensity Modulated Proton Therapy in Breath-hold for Locally Advanced Non-Small Cell Lung Cancer

International Journal of

Radiation Oncology biology

physics

www.redjournal.org

Physics Contribution

Feasibility of Pencil Beam Scanned Intensity Modulated Proton Therapy in Breath-hold for Locally Advanced Non-Small Cell Lung Cancer Jenny Gorgisyan, MSc,*,y,z Per Munck af Rosenschold, PhD,y,z Rosalind Perrin, PhD,* Gitte F. Persson, MD,y Mirjana Josipovic, MSc,y,z Maria Francesca Belosi, MSc,* Svend Aage Engelholm, MD,y Damien C. Weber, MD,*,x and Antony J. Lomax, PhD*,k *Center for Proton Therapy, Paul Scherrer Institute, Villigen PSI, Switzerland; yDepartment of Oncology, Rigshospitalet Copenhagen University Hospital, Copenhagen, Denmark; zNiels Bohr Institute, University of Copenhagen, Copenhagen, Denmark; xDepartment of Radiation Oncology, University Hospital of Zu¨rich, Zu¨rich, Switzerland; and kDepartment of Physics, ETH Zu¨rich, Zu¨rich, Switzerland Received May 18, 2017, and in revised form Jul 13, 2017. Accepted for publication Aug 16, 2017.

Summary The breath-hold approach as a motion mitigation technique in pencil beam scanning proton therapy was investigated with respect to intra- and interfraction residual motion. Fifteen patients with repeated breath-hold computed tomography scans were studied, and dose delivery was simulated, including patient motion and dynamic

Purpose: We evaluated the feasibility of treating patients with locally advanced non-small cell lung cancer (NSCLC) with pencil beam scanned intensity modulated proton therapy (IMPT) in breath-hold. Methods and Materials: Fifteen NSCLC patients who had previously received 66 Gy in 33 fractions with image guided photon radiation therapy were included in the present simulation study. In addition to a planning breath-hold computed tomography (CT) scan before the treatment start, a median of 6 (range 3-9) breath-hold CT scans per patient were acquired prospectively throughout the radiation therapy course. Three-field IMPT plans were constructed using the planning breath-hold CT scan, and the four-dimensional dose distributions were simulated, with consideration of both patient intra- and interfraction motion, in addition to dynamic treatment delivery. Results: The median clinical target volume receiving 95% of the prescribed dose was 99.8% and 99.7% for the planned and simulated dose distributions, respectively. For 3 patients (20%), the dose degradation was >5%, and plan adjustment was needed. Dose

Reprint requests to: Jenny Gorgisyan, MSc, WBBA/007 Paul Scherrer Institute, 5232 Villigen PSI, Switzerland. Tel: þ41 56 3105587; E-mail: [email protected] A preliminary analysis of the our study was presented at the Particle Therapy Co-operative Group (PTCOG) 2016, Prague, Czech Republic, and American Association of Physicists in Medicine (AAPM), 2016, Washington DC. Conflict of interest: J. Gorgisyan reports personal fees from the Danish Society for Clinical Oncology, during the conduct of the study. M. Josipovic reports grants from Varian Medical Systems, personal fees from Int J Radiation Oncol Biol Phys, Vol. 99, No. 5, pp. 1121e1128, 2017 0360-3016/$ - see front matter Ó 2017 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ijrobp.2017.08.023

Brain Lab, personal fees from Varian Medical Systems, outside the submitted work. Supplementary material for this article can be found at www.redjournal.org. AcknowledgmentsdThe authors thank Drs Marta Peroni and Ye Zhang for fruitful discussions regarding Plastimatch deformable image registration and the in-house 4D dose calculation software, Dr David Oxley for extensive programming support, and Lorentzos Mikroutsikos for valuable treatment planning advice. Furthermore, the authors thank the Danish Society for Clinical Oncology for financial support.

International Journal of Radiation Oncology  Biology  Physics

1122 Gorgisyan et al. treatment delivery using 4-dimensional dose calculations. Treatment delivery was feasible, and the plans were dosimetrically robust for 9 of 15 cases.

degradation correlated significantly with the change in water-equivalent path lengths (P<.01) in terms of the percentage of voxels with 3-mm or more undershoot on repeat CT scans. The dose to the organs at risk was similar for the planned and simulated dose distributions. Three or fewer breath-holds per field would be required for 12 of the 15 patients, which was clinically feasible. Conclusions: For 9 of 15 NSCLC patients, IMPT in breath-hold was both dosimetrically robust and feasible to deliver regarding the treatment time. Three patients would have required plan adaption to meet the dosimetric criteria. The change in water-equivalent path length is an indicator of plan robustness and should be considered for the selection of patients for whom the plan would require adaptation. Ó 2017 Elsevier Inc. All rights reserved.

Introduction Motion, for example due to breathing, can jeopardize the precision of proton therapy. Motion can lead to dose blurring and can interfere with dynamic treatment deliveries such as pencil beam scanned (PBS) proton therapy and lead to unwanted hot and cold spots in the delivered dose distribution (1, 2). Currently, many approaches are available to mitigate the intrafraction motion effect in proton therapy; for example, by increasing the margins of the target based on patient-specific tumor motion as quantified from four-dimensional CT (4D-CT) scans (3), through tumor tracking (4) or gating that accounts for motion (5), or by reducing the motion through abdominal compression (6) or forced or voluntary breath-holds (7, 8). Breath-hold has been extensively investigated for photon radiation therapy (9-11) but has remained relatively unassessed for proton therapy, with a few exceptions, as follows. Stuschke et al (8) concluded that treatment in a single breathhold combined with rescanning, would be a safe mode to treat moving targets using proton therapy. In addition, previous studies from our group have found that the breath-hold approach is robust to interfraction motion for early-stage lung cancer (7) and that the difference in water-equivalent path lengths (WEPLs), together with baseline shifts of the tumor, are good indicators of plan robustness in breath-hold treatment plans (12). Thus, investigation of the breath-hold approach for proton therapy is warranted, because proton therapy has greater potential compared with photon radiation therapy to escalate the dose to the tumor and reduce the dose to the lungs and heart (13). However, whether this can be translated to overall survival and toxicity benefits for lung cancer patients remains to be demonstrated (14). The breath-hold approach relies on patient compliance. As such, patients are required to hold their breath for up to 20 seconds repeatedly during treatment. Nonetheless, breath-hold has been shown to be well tolerated by patients (9, 10, 15), with good intrafraction (16) and interfraction (17) reproducibility. The aim of the present study was to explore the feasibility of breath-hold for PBS proton therapy for locally advanced non-small cell lung cancer (NCSLC) through simulations. Specifically, the dosimetric robustness to intra-

and interfraction motion was studied, together with the delivery feasibility of the generated proton plans, in terms of treatment time and machine delivery aspects. To the best of our knowledge, ours is the first investigation of the intraand interfraction breath-hold reproducibility and feasibility of treatment delivery for locally advanced NSCLC patients using PBS proton therapy.

Methods and Materials Patients and image data Breath-hold CT data from 17 patients were acquired at our institution (photon clinic) during 2012 to 2013. All patients were treated for locally advanced NSCLC with photon radiation therapy with 66 Gy in 33 fractions within 6 to 7 weeks (17). The patients (median age 67 years, range 46-80) were prospectively enrolled in a clinical protocol at the photon clinic (local ethics committee approval no. H-2-2011-153) and were included in the present retrospective PBS proton therapy simulation study (Danish Data Protection Agency j.nr. 2012-58-0004). A summary of the NSCLC characteristics of the patients is given in Table 1. Each patient underwent a breath-hold CT scan at the time of planning, and 3 breath-hold CT scans during 3 consecutive breath-holds at days 2 (CT2), 16 (CT16), and 31 (CT31), total 10 breath-hold CT scans per patient. The visually guided voluntary deepinspiration breath-holds (coached before treatment), controlled using the Real-Time Position Management System (Varian Medical Systems Inc, Palo Alto, CA), were held on average for 19  6 seconds (mean  standard deviation), and the duration between the breath-holds was 38  11 seconds. Of the approached patients, 10% declined study participation, with an inability to hold the breath as the reason; w50% declined participation in total. Of the 17 patients enrolled, two patients were excluded from the study because of data loss of the planning CT scan. For various reasons, a number of repeated CT scans were missing for the remaining 15 patients (Table E1; available online at www .redjournal.org), resulting in a missing CT scan rate of 22%. The missing data were replaced using the closest acquired CT scan (eg, CT31 was replaced with CT16).

Volume 99  Number 5  2017 Table 1 patients

Breath-hold feasibility for PBS IMPT 1123

Nonesmall cell lung cancer characteristics of

Pt. no.

NSCLC stage

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

T3N2M0 T2BN3M0 T3N2M0 T3N0M0 T4N3 T4N2M0 T4N2M0 T2BN2M0 T4N2M0 T4N2 T4N2M0 T4N0M1B* TXN3M0 T1BN2M0 TXN3M0

CTV (cm3) 371 183 142 130 82 509 199 144 958 145 148 66 101 75 174

Target localization Right lobe Left lobe Left lobe Right lobe Right lobe Left lobe Left lobe Right lobe Right lobe Left lobe Right lobe Left lobe Right lobe Right lobe Right lobe

Abbreviations: CTV Z clinical target volume; NSCLC Z non-small cell lung cancer; Pt. no. Z patient number. * Cerebral metastasis treated with stereotactic radiosurgery.

perpendicular to the beam direction. For 1 patient (patient 9), the separations were increased to 6 mm in the latter axes for technical reasons related to our in-house treatment planning system. The treatment plans were initially normalized to the mean PTV dose. The normalization was then increased 1% to 3% to maintain the PTV receiving 95% of the prescribed dose (V95%) at w95%. All treatment plans were approved by a radiation oncologist with specialization in lung cancer radiation therapy. The dose constraints were met in all cases, except for the esophagus D2% for 4 patients (27%) [median 1.4 Gy(RBE), range 0.3-2.7 Gy(RBE), greater than the dose constraint of 66 Gy(RBE)]. Clinical pretreatment verification was performed for 1 example patient, for whom the measured 2-dimensional dose profiles presented a dose deviation within 2% of the nominal dose. The 3-dimensional dose distribution reconstructed using the machine log file information (22, 23) deviated <1% from the nominal dose. Both results were in line with the clinical cases treated at our institute.

Simulations Treatment planning A single radiation oncologist delineated the gross target volume (GTV) and clinical target volume (CTV). The mean CTV was 224  221 cm3 (Table 1). The CTV to planning target volume (PTV) margin was calculated according to van Herk et al (18) using clinical breath-hold CT data acquired at our institution (17) and an estimation of the proton beam penumbra of 8 mm. The PTV margins were calculated to be 5 mm in the anteroposterior and craniocaudal directions and 4 mm in the left-right direction. Intensity modulated proton therapy (IMPT) treatment plans using multifield optimization (19) with 3 fields each were designed using the breath-hold planning CT scan in the in-house treatment planning system used clinically at our institute (20). The relative biologic effectiveness (RBE) was assumed to be a constant value of 1.1, the standard practice at our institute and in line with other studies (21). The aim was to achieve a homogenous dose of 66 Gy(RBE) in 33 fractions to the target using organs at risk (OAR) constraints as follows, with DX% the dose to X% of the volume and VYGy, the volume that received Y Gy:  Spinal cord: D2% 45 Gy(RBE)  Both lungs (minus the GTV): V20Gy 37%, V5Gy 60%, and mean dose 20 Gy(RBE)  Heart: V50Gy 20% and mean dose 45 Gy(RBE)  Esophagus: D2% 66 Gy(RBE) A robust beam arrangement was selected such that each beam would only traverse regions presenting small anatomic differences between the planning and repeated breath-hold CT scans (CT2 data set). The Bragg peak separations (and voxel size of the dose calculation) were 2.5 mm along the beam axis (energy separation) and 4 mm along the axes

The dose for the entire treatment was subsequently simulated using our 4D dose calculation software based on the deformed dose grid approach (1, 24), and adapted to use with the repeated breath-hold CT data. The dose calculation itself was performed using the ray-casting technique (25) and our in-house treatment planning system (20). Time stamps were calculated for each field according to the time used for delivering each spot and the associated dead times for energy changes and magnetic scanning (24). The sum of these time steps gave the overall treatment time. The amount of intrafraction breath-hold CT scans was determined from the overall treatment time, assuming w20 seconds per breath-hold (17) (eg, 60 seconds of treatment time corresponded to 3 breath-hold CT scans), indicating that the spots were simulated to be delivered in different geometries. From clinical experience at our institution (photon clinic), treatment delivery was deemed feasible if 3 breath-holds per field were required. The deformed dose grid approach deforms the dose calculation grid according to the motion vectors originating from deformable image registrations (DIRs) between the CT scans (Fig. 1). DIR was performed in Plastimatch (available at: http://www.plastimatch.org), first with respect to the entire image and, subsequently, the area of the PTV plus 3 cm, with the resulting vector fields used as input for the 4D dose calculations. Rigid image registration was performed before DIR for patient setup according to the target alignment. The DIR parameters were determined from the work by Sharp et al (26) and optimized for the present data set. The images were registered using both B-spline and Demons algorithms to investigate the sensitivity of the results to the DIR algorithm. Using this approach, the doses calculated on the repeat CT scans representing different breath-hold scenarios during the

International Journal of Radiation Oncology  Biology  Physics

1124 Gorgisyan et al.

B

Planning CT

Repeated CT

Deformable image registration

Vector fields

+

Dose grid

Inter-fractional set of CT scans

A

Intra-fractional CT scans

Fraction 1

CT2_2

CT2_2

CT2_1

Fraction 2 . . .

CT2_3

CT2_2

CT2_1

Fraction 12 . . .

CT16_1

CT16_1

CT16_3

Fraction 33

CT31_1

CT31_2

CT31_3

=

Deformed dose grid

Fig. 1. (A) The original dose grid was deformed by the vector fields arising from the deformable image registration between the planning and repeated computed tomography (CT) scans and subsequently used for four-dimensional dose calculation. (B) Example of the distribution of the CT scans for the simulation (treatment time 60 seconds). From the 9 available repeated breath-hold CT scans from 3 different days (day X: CTX_1, CTX_2, CTX_3), the 3 data sets were distributed over the fractions as follows: fractions 1 to 11, day 2; fractions 12 to 22, day 16; and fractions 23 to 33, day 31. The choice of the intrafraction breath-hold CT scans was random. treatment course were accumulated on the original planning CT scan for comparison with the original treatment plan. It was assumed that fractions 1 to 11 could be represented by the CT2 data set, fractions 12 to 22 by the CT16 data set, and so forth, and the choice of the intrafraction breath-hold CT scans was random (Fig. 1). Therefore, each intrafraction breath-hold CT scan and the corresponding dose grid could be used multiple times, or not at all, during the simulation of 1 treatment field. The fields and the respective dose matrices were subsequently summed to give the total dose for 1 single fraction. The final dose distribution was achieved by summing the 33 fractions. Overall, 1485 (3 fields, 33 fractions, 15 patients) 4D dose distributions were computed.

Armonk, NY), with a significance level of PZ.05. The following parameters were assessed for correlation: baseline shift of the tumor (17), change in lung volume (17), and changes in WEPL. The latter was quantified using the mean WEPL difference, the percentage of voxels with >3 mm of over- or undershoot, and the V95% of the WEPL over- or undershoot distributions, as suggested by Veiga et al (27). In the present study, the WEPL was computed as the path length to the distal edge voxels of the target (ie, anatomic changes both within and proximal to the target will be detected), multiplied by the relative stopping power to water of protons in the traversed material (12).

Evaluation

The relative dose differences between the planned and simulated treatments were, on average, for all patients, a maximum of 17% inside the CTV but 52% outside the target (Fig. 2). However, because the dose outside the target was low, the dose constraints were met in almost all cases. The exceptions were the lung V20Gy for patient 9, for whom the dose was 0.2% greater than the constraint (37%) and esophagus D2% for 8 of 15 patients, with the dose in the worst case 5.1 Gy(RBE) greater than the constraint [66 Gy(RBE)]. The median CTV V95% for all patients was 99.8% (range 99.5%-100%) and 99.7% (range 73.7%-100%) for the nominal and simulated plans, respectively (comparison

The dose distributions and dose-volume histograms were analyzed using MATLAB, version R2016a (MathWorks, Natick, MA). The parameters, including the V95% and the difference between the percentage of the prescribed dose received by 5% and 95% of the target (D5%  D95%) for the CTV, together with the relevant dose metrics for OARs, were evaluated. Degradation of the CTV V95% to <5% was deemed acceptable. Wilcoxon signed rank tests for differences and mixed linear model tests for correlations were performed using SPSS Statistics, version 22 (IBM Corp,

Results

Volume 99  Number 5  2017

Breath-hold feasibility for PBS IMPT 1125

Treatment plan

Cumulative dose distribution

Difference

Patient 1

Patient 2

Patient 3

Patient 4

Patient 5

Patient 6

Patient 7

Patient 8

Treatment plan

Cumulative dose distribution

Difference

Dose

Patient 9

Diff

Patient 10 Patient 11 Patient 12 Patient 13 Patient 14 Patient 15

Fig. 2. The dose distributions in color wash (Dose) for the (Top) treatment plan and (Middle) simulated cumulative dose distributions, together with (Bottom) voxel-by-voxel relative dose difference distributions in color wash (Diff). White lines indicate the clinical target volume. The dose is shown as the percentage of the prescribed dose [66 Gy(RBE)]. between nominal and simulated dose values: PZ.05; Fig. 3). The corresponding median CTV D5%  D95% was 5.4% (range 3.0%-7.3%) and 7.5% (range 4.9%-20.1%; P<.01). The median lung (PZ.02) and heart (PZ.21) mean dose CTV Coverage

Dose to organs at risk 70

90 80 70

Simulated CTV Homogeneity

50 40 30 20 10 0

Planned

Simulated

Lu n

25 20 15 10 5 0

Planned

Dose [Gy(RBE)]

60

g m Lu ean ng pl a m ea n He n ar t m sim ea n He pl ar an tm Sp in e an al c sim Sp ord in D2 al % co pl an rd Es op D2 ha % gu s s D im Es 2% op pl ha an gu sD 2% sim

V95% [%]

100

D5%-D95% [%]

were 10 Gy(RBE), and the median spinal cord D2% (PZ.07) was 5 Gy(RBE). The median esophagus D2% for the nominal and simulated plans was 65.3 Gy(RBE) and 66.3 Gy(RBE), respectively (PZ.01; Fig. 3).

Fig. 3. (Left) The planned (plan) and simulated (sim) dose to the clinical target volume (CTV) are shown as percentage of volume receiving 95% of the prescribed dose (V95%) and difference between the percentage of the prescribed dose received by 5% and 95% of the volume (D5%  D95%). (Right) The mean dose to the heart and lung, together with D2% for the spinal cord and esophagus. In the boxplots, the horizontal red lines indicate the median values, the blue boxes, the 25th to 75th quartiles, the vertical black dotted lines, the ranges, and the red crosses, the outliers (outside 2.7s) of the data distribution.

1126 Gorgisyan et al.

International Journal of Radiation Oncology  Biology  Physics

Of the 15 patients, 10 had a CTV V95% difference of <1%, and 12 had a difference of <5% (Fig. 4). Six patients had >3 mm WEPL undershoot in 15% of the voxels. Three of these had a >5% difference in the V95%. These patients presented with anatomic differences related to breath-hold reproducibility alone (nZ1) or combined with setup errors (nZ2). Degradation in CTV V95% was significantly associated with the percentage of voxels with >3 mm undershoot (P<.01) using linear models. This was the only statistically significant correlation found between the degradation in dose and the studied parameters. The median estimated treatment delivery time was 58.3 seconds per field (range 28.2-181.1), equivalent to a median of 3 breath-holds per field (range 1-9; Fig. 4). Of the 15 patients, 12 (80%) would require 3 breath-holds per field for treatment delivery. The remaining 3 patients, with a target volume >200 cm3, would have required 4 breath-holds (patient 1), 6 breath-holds (patient 6), and 9 breath-holds (patient 9) per field. The treatment time correlated with the size of the CTV (P<.01). Differences in dose when performing DIR with the B-spline versus Demons algorithms were low (Table E2; available online at www.redjournal.org).

failed using our criteria. Szeto et al (29) showed that 8 of 16 locally advanced NSCLC PBS proton therapy treatment plans were not robust using their criterion of 2 Gy(RBE) underdosage of the GTV D99%. With that criterion, 3 of our plans would also have failed; however, only 2 were among the same that failed using our criteria. Both of these studies investigated free-breathing image data, in contrast to the breath-hold image data evaluated in the present study. A possible explanation could be the interfraction reproducibility between the mid-ventilation 4D-CT phases (28) or that of the cone beam CT scans (29) might result in lower geometric reproducibility than that betweenebreath-hold CT scans. For free-breathing photon radiation therapy, the rate of replanning has been shown to be w30% (30). As an alternative to replanning, robust treatment planning (8, 31, 32) can be used, for which, in addition to setup and range errors, the motion can be included (33); for example, through population-based lung models (34). Although this method can increase the robustness of the treatment plan, it might be at the expense of the initial plan quality. For some of our plans, the dose constraint for the esophagus could not be met. Because the esophagus was located inside the PTV (nZ13), or in close proximity to the PTV (nZ2), the dose was inevitably high to ensure the target coverage. Chang et al (14) reported that grade 2 and 3 esophagitis developed in 30% and 10% of patients, respectively, who had undergone passively scattered proton therapy 74 Gy(RBE) with concurrent chemotherapy. In contrast, the reported esophagus mean dose of 21.2 Gy(RBE) and V50Gy of 20.4% were greater than the doses in the present study of 14.4 Gy(RBE) and 19.2%, respectively. The differences in the WEPL appeared to be an important predictor of plan robustness (Fig. 4), confirming previous results (12). WEPL calculation is faster than full dose calculation and has, therefore, a potential to be used online. The evaluation of WEPL, using daily CT imaging,

Discussion Dose degradation to the target and the dose increase to the OARs due to inter- and intrafraction residual breath-hold motion was found to be minor for most patients (Figs. 2-4). As shown, 12 of 15 patients had a CTV V95% degradation of <5%, indicating good plan robustness. This was in contrast to the findings reported by Hoffmann et al (28), who showed that plan adaption was necessary for most locally-advanced NSCLC cases treated with IMPT. With their criterion of robustness (CTV V95% >95%), 3 of our plans would have failed, the same ones that would have Dose degradation as a function of WEPL difference PT1 PT2 PT3 PT4 PT5 PT6 PT7 PT8 PT9 PT10 PT11 PT12 PT13 PT14 PT15

100 95 90 85 80

Delivery time

200 180

PT1 PT2 PT3 PT4 PT5 PT6 PT7 PT8 PT9 PT10 PT11 PT12 PT13 PT14 PT15

160 Time per field [s]

CTV V95% percentage ratio

105

140 120 100 80 60

75 70

40 0

10 20 30 40 50 Voxels with > 3mm undershoot of the beam [%]

20

0

200

400 600 CTV volume [cm3]

800

1000

Fig. 4. (Left) The percentage of the clinical target volume (CTV) receiving 95% of the prescribed dose (V95%) ratios between the original and simulated plans as a function of the water-equivalent path length (WEPL) difference. The black dotted line indicates a 5% decrease in the CTV V95%. (Right) The mean treatment delivery time per field as a function of the CTV size. The black dotted line indicates 3 breath-holds per field. Abbreviation: PT Z patient.

Volume 99  Number 5  2017

combined with visual evaluation of the daily verification imaging could potentially lead to more clinical confidence in dose delivery for mobile lung cancers. Residual motion during the breath-holds was not considered in the present study but will be the subject of further studies by our group. However, residual motion of the target during breath-hold has been reported to be as low as 1.4, 1.2, and 2.1 mm for the anteroposterior, left-right, and craniocaudal directions, respectively (16). An increased amount of CT data could have increased the validity of the present study but would have resulted in a greater image radiation dose to the patient. Another approach could be to use virtual CT scans simulated from cone beam CT scans registered to the planning CT scan (27). A key aspect of our simulations was the use of DIR. DIR is an important, but controversial, tool in radiation therapy when evaluating the dose to the patient throughout the treatment course. Uncertainties related to DIRs of the lung region range from the submillimeter range to 3 mm (35). As a sensitivity measure in the present study, the 4D dose calculations were performed using both B-spline and Demons DIR algorithms. Our results revealed that the only statistically significant change was in the mean lung dose (Table E2; available online at www.redjournal.org), but this was very small [0.11 Gy(RBE)] and not clinically significant. These results enforce the reliability of the performed DIR. After visual inspection of the DIRs, the Jacobian matrices of the vector fields were computed to ensure a lack of unrealistic organ deformations such as folding of the organs (negative Jacobian) (36). The only patient not complying with this restriction (patient 9) had presented with both a large tumor and large tumor shrinkage. Motion restriction was attempted but resulted in poorer agreement between the fixed and warped images. Additional dose simulations with both scenarios revealed differences in the V95% within 1%, indicating the low sensitivity of this parameter to the subsequent dose simulations. However, it remains unclear whether the DIR algorithm can account for the large tumor shrinkage accurately, leading to uncertainties in the resulting dose distribution for this particular patient. Finally, the calculated delivery times indicated that 12 of 15 patients, with tumor sizes <200 cm3 (Table 1 and Fig. 4), would require 3 breath-holds per field. The overall treatment time, in addition to the beam delivery time, would also include the duration of the pause between breath-holds and the time required for patient setup. Overall, 9 of 15 plans were shown to be both dosimetrically robust and feasible to deliver in terms of the required amount of breath-holds. For the 3 patients requiring the longest treatment times, 9 breathholds per field were necessary, which is clinically unfeasible. As such, improvements in the treatment time would be warranted for these patients before implementing the breathhold technique clinically. For instance, the treatment time could be decreased by optimization of delivery parameters such as the spot distance or scanning speed; however, this could also alter the robustness of the plan. At our institute, it

Breath-hold feasibility for PBS IMPT 1127

is also possible to deliver fields using the so-called line scanning technique, as another example, in which the dose is delivered as continuous lines instead of as discrete pencil beams (37). This delivery technique is currently at the experimental stage; however, promising results have been demonstrated regarding substantial decreases in treatment time (38).

Conclusions Considering inter- and intrafraction breath-hold reproducibility, the breath-hold approach has been shown to be robust for IMPT for locally advanced NSCLC. Delivery was possible using 3 breath-holds per field for 12 of 15 patients, which is clinically feasible. Patients with large tumors >200 cm3 required unfeasible treatment times; thus, improved treatment delivery speed would be required (eg, using line scanning). Overall, for 9 of 15 patients (60%), proton therapy was deemed to be both dosimetrically robust and deliverable using 3 breath-holds. The results showed <5% degradation in CTV V95% for 12 of 15 patients throughout the treatment course, and the difference in WEPL, as calculated on different CT scans, was a good indicator of plan robustness.

References 1. Zhang Y, Boye D, Tanner C, et al. Respiratory liver motion estimation and its effect on scanned proton beam therapy. Phys Med Biol 2012; 57:1779-1795. 2. Grassberger C, Dowdell S, Lomax A, et al. Motion interplay as a function of patient parameters and spot size in spot scanning proton therapy for lung cancer. Int J Radiat Oncol Biol Phys 2013;86:380386. 3. Dowdell S, Grassberger C, Sharp GC, et al. Interplay effects in proton scanning for lung: A 4D Monte Carlo study assessing the impact of tumor and beam delivery parameters. Phys Med Biol 2013;58:41374156. 4. Rosenscho¨ld PMA, Aznar MC, Nygaard DE, et al. A treatment planning study of the potential of geometrical tracking for intensity modulated proton therapy of lung cancer. Acta Oncol 2010;49:11411148. 5. Mori S, Wolfgang J, Lu HM, et al. Quantitative assessment of range fluctuations in charged particle lung irradiation. Int J Radiat Oncol Biol Phys 2008;70:253-261. 6. Lin L, Souris K, Kang M, et al. Evaluation of motion mitigation using abdominal compression in the clinical implementation of pencil beam scanning proton therapy of liver tumors. Med Phys 2017;44:703-712. 7. Dueck J, Knopf A-C, Lomax A, et al. Robustness of the voluntary breath-hold approach for the treatment of peripheral lung tumors using hypofractionated pencil beam scanning proton therapy. Int J Radiat Oncol Biol Phys 2016;95:534-541. 8. Stuschke M, Kaiser A, Po¨ttgen C, et al. Potentials of robust intensity modulated scanning proton plans for locally advanced lung cancer in comparison to intensity modulated photon plans. Radiother Oncol 2012;104:45-51. 9. Hanley J, Debois MM, Mah D, et al. Deep inspiration breath-hold technique for lung tumors: The potential value of target immobilization and reduced lung density in dose escalation. Int J Radiat Oncol Biol Phys 1999;45:603-611.

1128 Gorgisyan et al.

International Journal of Radiation Oncology  Biology  Physics

10. Josipovic M, Persson GF, Ha˚kansson K, et al. Deep inspiration breath hold radiotherapy for locally advanced lung cancer: Comparison of different treatment techniques on target coverage, lung dose and treatment delivery time. Acta Oncol 2013;52:1582-1586. 11. Marchand V, Zefkili S, Desrousseaux J, et al. Dosimetric comparison of free-breathing and deep inspiration breath-hold radiotherapy for lung cancer. Strahlenther Onkol 2012;188:582-591. 12. Gorgisyan J, Perrin R, Lomax AJ, et al. Impact of beam angle choice on pencil beam scanning breath-hold proton therapy for lung lesions. Acta Oncol 2017;56:853-859. 13. Chang JY, Li H, Zhu XR, et al. Clinical implementation of intensity modulated proton therapy for thoracic malignancies. Int J Radiat Oncol Biol Phys 2014;90:809-818. 14. Chang JY, Komaki R, Lu C, et al. Phase 2 study of high-dose proton therapy with concurrent chemotherapy for unresectable stage III nonsmall cell lung cancer. Cancer 2011;117:4707-4713. 15. Panakis N, McNair HA, Christian JA, et al. Defining the margins in the radical radiotherapy of non-small cell lung cancer (NSCLC) with active breathing control (ABC) and the effect on physical lung parameters. Radiother Oncol 2008;87:65-73. 16. Rydho¨g JS, de Blanck SR, Josipovic M, et al. Target position uncertainty during visually guided deep-inspiration breath-hold radiotherapy in locally advanced lung cancer. Radiother Oncol 2017;123: 78-84. 17. Josipovic M, Persson GF, Dueck J, et al. Geometric uncertainties in voluntary deep inspiration breath hold radiotherapy for locally advanced lung cancer. Radiother Oncol 2016;118:510-514. 18. van Herk M, Remeijer P, Rasch C, et al. The probability of correct target dosage: Dose-population histograms for deriving treatment margins in radiotherapy. Int J Radiat Oncol Biol Phys 2000;47: 1121-1135. 19. Lomax AJ. Intensity-Modulated Proton Therapy. In: DeLaney TF, Kooy HM, editors. Proton and charged particle radiotherapy, 1st ed. Philadelphia: Wolters Kluwer, Lippincott Wiliams & Wilkins; 2008. pp. 98-107. 20. Lomax AJ, Bo¨hringer T, Bolsi A, et al. Treatment planning and verification of proton therapy using spot scanning: Initial experiences. Med Phys 2004;31:3150-3157. 21. Paganetti H. Relative biological effectiveness (RBE) values for proton beam therapy: Variations as a function of biological endpoint, dose, and linear energy transfer. Phys Med Biol 2014;59: R419-R472. 22. Scandurra D, Albertini F, Van der Meer R, et al. Assessing the quality of proton PBS treatment delivery using machine log files: Comprehensive analysis of clinical treatments delivered at PSI Gantry 2. Phys Med Biol 2016;61:1171-1181. 23. Meier G, Besson R, Nanz A, et al. Independent dose calculations for commissioning, quality assurance and dose reconstruction of PBS proton therapy. Phys Med Biol 2015;60:2819-2836.

24. Boye D, Lomax T, Knopf A. Mapping motion from 4D-MRI to 3D-CT for use in 4D dose calculations: A technical feasibility study. Med Phys 2013;40:061702. 25. Schaffner B, Pedroni E, Lomax A. Dose calculation models for proton treatment planning using a dynamic beam delivery system: An attempt to include density heterogeneity effects in the analytical dose calculation. Phys Med Biol 1999;44:27-41. 26. Sharp G, Peroni M, Li R, et al. Evaluation of Plastimatch B-spline registration on the EMPIRE10 data set. In: Medical Image Analysis for the Clinic: A Grand Challenge (in conjunction with MICCAI 2010). Beijing, China: MICCAI; 2010. p. 99-108. 27. Veiga C, Janssens G, Teng CL, et al. First clinical investigation of cone beam computed tomography and deformable registration for adaptive proton therapy for lung cancer. Int J Radiat Oncol Biol Phys 2016;95:549-559. 28. Hoffmann L, Alber M, Jensen MF, et al. Adaptation is mandatory for intensity modulated proton therapy of advanced lung cancer to ensure target coverage. Radiother Oncol 2016;122:400-405. 29. Szeto YZ, Witte MG, van Kranen SR, et al. Effects of anatomical changes on pencil beam scanning proton plans in locally advanced NSCLC patients. Radiother Oncol 2016;120:286-292. 30. Møller DS, Holt MI, Alber M, et al. Adaptive radiotherapy for advanced lung cancer ensures target coverage and decreases lung dose. Radiother Oncol 2016;121:32-38. 31. Liu W, Liao Z, Schild SE, et al. Impact of respiratory motion on worstcase-scenario optimized intensity-modulated proton therapy for lung cancers. Pract Radiat Oncol 2015;5:e77-e86. 32. Unkelbach J, Chan TC, Bortfeld T. Accounting for range uncertainties in the optimization of intensity modulated proton therapy. Phys Med Biol 2007;52:2755-2773. 33. Bernatowicz K, Zhang Y, Perrin R, et al. Advanced treatment planning using direct 4D optimisation for pencil-beam scanned particle therapy. Phys Med Biol 2017 [Epub ahead of print]. https://doi.org/10.1088/ 1361-6560/aa7ab8. 34. Szeto YZ, Witte MG, van Herk M, et al. A population based statistical model for daily geometric variations in the thorax. Radiother Oncol 2017;123:99-105. 35. Brock KK, Velec M, Lee JH. Validation of image registration. In: Brock KK, editor. Image Processing in Radiation Therapy. Boca Raton, FL: CRC Press; 2014. p. 41-62. 36. Varadhan R, Karangelis G, Krishnan K, et al. A framework for deformable image registration validation in radiotherapy clinical applications. J Appl Clin Med Phys 2013;14:192-213. 37. Zenklusen SM, Pedroni E, Meer D. A study on repainting strategies for treating moderately moving targets with proton pencil beam scanning at the new Gantry 2 at PSI. Phys Med Biol 2010;55:5103-5121. 38. Scha¨tti A, Meer D, Lomax AJ. First experimental results of motion mitigation by continuous line scanning of protons. Phys Med Biol 2014;59:5707-5723.