53rd SSRMP Annual Meeting

21 Nov 2019, 09:00 → 22 Nov 2019, 18:00 Europe/Zurich

Paul Scherrer Institute

We are looking forward to welcome you to the 53rd SSRMP Annual Meeting at the Paul Scherrer Institute in Villigen. Thanks to your contribution, an attractive and engaging scientific program will be prepared. A lovely social event will take place at the FHNW Campus in Windisch with the opportunity to practice ice-skating for the adventurers. We are also delighted to offer you a guided tour of the Center of Proton Therapy at PSI.
Finally, but not least, the program will be also enriched with keynotes from three special guests:

Prof Thomas Bortfeld
Harvard Medical School and Massachusetts General Hospital, Boston, USA

Prof Rob Coppes
University Medical Center and University of Groningen, Groningen, The Netherlands

Prof Günther Dissertori
Institute for Particle Physics and Astrophysics, ETH Zürich, Zürich, Switzerland

Abstract deadline: 30th August extended to 13th September

Registration deadline: 25th October extended to 10th November

Social event: Thursday evening 21st of November

Tour of the Proton Therapy Center: late afternoon Friday 22nd of November

We therefore encourage you to register for the event and invite you to submit your abstract as soon as possible.

For the organizing committee

Sairos Safai, Francesca Belosi and Tony Lomax

 

Thursday 21 November

09:00 → 10:00 Registration

10:00 → 10:15 Opening Remarks

10:15 → 11:30 Session I: Image Guidance - Chair: Prof. Philippe Cattin

Department of Biomedical Engineering, University of Basel

10:15 Targeting arteriovenous malformations without an invasive head ring - A new software prototype

Introduction
Arteriovenous malformations (AVMs) are abnormal, snarled tangles of blood vessels that cause multiple irregular connections between the arteries and veins. In order to correctly locate an AVM, typically, a digital subtraction angiography (DSA) is carried out. To use the DSA for target definition an accurate image registration between CT and DSA is required. Carrying out a non-invasive, frameless procedure, i.e. having no invasive head frame in place that serves as a fixed coordinate system, registration of the 2D-DSA images with the CT is critical. A new software prototype (called prototype below) is enabling this frameless procedure. The aim of this work was to evaluate the prototype in terms of targeting accuracy and reliability based on phantom measurements as well as with the aid of patient data. In addition, the user's ability to recognize mismatches on the present image modalities and quality was assessed.

Materials and Methods
Targeting accuracy was measured with a simple cubic, as well as with a more realistic anthropomorphic head phantom. Clearly defined academic targets within the two phantoms were contoured on the CT. The center of mass (COM) coordinates of these reference structures were compared with the COM coordinates of the structures generated within the prototype, based on the 2D-DSA images. A similar approach was used with patient data, where the clinically contoured target served as the reference structure.
In order to check if the user is able to recognize registration errors on blended digitally reconstructed radiographs (DRR) and 2D-DSA projection images, a set of different registration errors (translations and rotations) was introduced to the correctly registered CT and 2D-DSA image data sets of three different patients. Each of six different users rated the whole set of registrations within the prototype.

Results
The target accuracy of the prototype was found to be below 0.04 cm for the simple cubic phantom and below 0.05 cm for the anthropomorphic head phantom. The mean target accuracy for the 15 patient cases was found to be below 0.3 cm.
Almost all introduced registration errors above 1° or 0.1 cm were detected by the six users and were rated as not acceptable. Nevertheless, in order to quantify and categorize the possibility to detect mismatches (sensitivity and specificity) in the registration process more data needs to be evaluated.

Conclusions
Our study shows, that the prototype is a useful tool that has the potential to fill the gap towards a frameless procedure when treating AVMs with the aid of 2D-DSA images in radiosurgery. The target accuracy of the prototype is similar to other systems, which are already well established in clinical routine.

Speaker: Daniel SCHMIDHALTER (Division of Medical Radiation Physics and Department of Radiation Oncology, Inselspital, Bern University Hospital, and University of Bern, Switzerland.)

10:30 Motion management and dose reconstruction for MRI guided cardiac radiosurgery

Purpose: To evaluate the dosimetric effect of the residual breathing motion during gating for the first patient treated with MRI guided cardiac radiosurgery.

Methods: A patient with ventricular tachycardia was treated in one fraction with a dose of 25 Gy to the 85%-isodose line to the planning target volume (PTV). The target volume (TV) was defined according to the cardiologists interpretation of the electrophysiological mapping of the arrhythmic substrat and expanded by 2 mm axially and 3 mm craniocaudally to the PTV. The patient was treated with the MRIdian system (Viewray). A mix of instructed breath-hold and free-breathing was performed, while a structure close to the diaphragm was tracked for respiratory gating at end of-exhale using sagittal 2D cine-MR with 4 fps. A gating window of 5 mm was defined and maximal 20% of the structure were allowed outside of the window for irradiation. The residual motion during gating was evaluated in the two directions (SI and AP) and a 2mm-binned position-probability map was generated. The dose to the patient was recalculated for each possible combination of SI- and AP-shift and summed with the weighting of the probability map.

Results: In total, 46 min of cine-MR were recorded for treatment and the target was within the gating window for 24 min. Thereof, 14 min were in breath-hold (>=10 s, 55 holds) and 10 min in free-breathing (<10 s, 257 breaths). The mean 2D offset could be reduced from 6.8 mm to 3.7 mm with gating. The reconstructed dose distribution showed a TV Dmean of 27.6 Gy (planned: 27.8 Gy) and D95 of 26.3 Gy (planned: 27 Gy). Organs at risk did not show any increase in dose, expect for the bronchial tree with a Dmean of 2.5 Gy (planned: 1.6 Gy).

Conclusion: MRI guided cardiac radiosurgery was successfully applied. With the motion tracking and gating, the patient could be treated with a good coverage of the target volume and minimal dose to organs at risk.

Speaker: Stefanie EHRBAR (Radiation Oncology Department, Universitätsspital Zürich, Switzerland)

10:45 Benefit of adaptation in MR-guided radiotherapy of lesions in the abdomen

Aims
The precise treatment of tumor lesions in the abdomen is challenging due to inter- and intra-fractional anatomy variations. With MR-guided radiotherapy, daily plan adaption can be performed. Here we investigated the benefit of adaption in terms of tumor coverage and organ at risk (OAR) sparing.

Methods
SBRT radiotherapy treatment plans, created with IMRT and treated with a gated breathhold technique on the MRIdian system (Viewray) for patients with cancer in the abdomen were analyzed. We chose 7 consecutive patients with the identical fractionation scheme (5 fractions, prescription 30-40Gy) and compared the delivered (adapted) doses with the predicted doses (original plans computed on the MRI of the day). In the MRIdian system, there are 2 types of plan re-optimization available: predefined segments weight optimization or full re-optimization, which can be done if weight optimization is not sufficient. We analyzed tumor coverage and OAR sparing using GTV D95%, PTV D95% and OAR V26Gy. The studied OAR was either stomach, duodenum or bowel, depending on the proximity to the PTV. For the OAR, we investigated for how many fractions V26Gy<1cc was exceeded, compared with the predicted dose. The statistical analysis was done with a paired-sample Wilcoxon test with a significance level <0.05. For all analyzed structures, the median difference and interquartile range were reported.

Results
The plan was re-optimized for all patients in all fractions. A full re-optimization was needed in at least 2 out of 5 fractions for each patient; for the remaining fractions weight optimization was performed and considered as sufficient based on the original plan. The GTV and PTV coverage were improved in 54% and 63% of the fractions, respectively. The OAR volume receiving 26Gy was reduced in 83% of the fractions. The V26Gy<1cc for OAR was exceeded in 77% of the predicted doses (57% original dose), whereas this number was reduced for the adapted doses (60%). For 2 patients, the GTV and PTV coverage were intentionally reduced in the adapted (and for 1 patient in the original) plan in order to improve OAR sparing. In the entire cohort, the OAR spearing was significantly improved: difference V26Gy=-1.05 (-4.20; -0.07) cc, p=0.01, whereas the tumor coverage was not significantly reduced: difference PTVD95% =0.46 (-0.50; 0.88) Gy, p=0.86; difference: GTVD95% =0.32 (-0.55; 1.44) Gy, p =0.86.

Conclusion
Results of this study show that MR-guided adaptive radiotherapy can notably reduce the dose to the organs at risk, while maintaining the GTV and PTV coverage in abdominal cancer.

Speaker: Ms Manon SPANIOL (Department of Radiation Oncology, University Hospital Zurich and University of Zurich, Zurich, Switzerland)

11:00 Time-Resolved 4DMRI

Introduction
4D imaging is an important tool for respiratory motion quantification and, among others, finds its application in radiotherapy treatment planning. To date, most 4D imaging techniques rely on prospective or retrospective stacking of partial image data and binning [1]. The result is a respiratory-correlated 4D image representing a mean breathing cycle. However, such binning methods lack the ability to reveal respiratory variabilities which are crucial for the validation of treatment plans or motion mitigation strategies in radiotherapy [2]. We present a novel 4D MRI approach which relies on a sophisticated k-space sampling strategy and motion compensation via spatial core alignment [3]. The presented method is truly time-resolved in the sense that it does not depend on respiratory binning but it provides a continuous 4D image while capturing motion variabilities.

Materials and Methods
The amount of time required for full Cartesian k-space sampling often exceeds the time frame where moving structure can be imaged without suffering from major motion artifacts in the final reconstruction. Thus, we have developed an MR sequence which alternately samples the k-space core and a peripheral patch of the k-space (Figure 1a). The acquired k-space cores allow the monitoring of the motion which happens during the acquisition of the peripheral patches. Peripheral patches are corrected for the motion by using the spatial image alignment of the cores. For each acquired core the spatially aligned peripheral patches are accumulated to a full and consistent time-resolved k-space [3].

Results
We modified a product sequence for a radio-frequency spoiled gradient echo acquisition with short repetition time and low flip angle in order to alter the sampling of the k-space. Considering 1500 time points, the acquisition time was 11.1 min and the reconstruction time 2h. Six volunteers have been scanned under free-breathing on a 3T clinical scanner. In Figure 1b, we compare sample slices of reconstructions where no motion has been considered (static) with the reconstructions where the non-rigid motion has been compensated.

Conclusions
By visually comparing the static and non-rigid reconstruction (Figure 1b), the motion-corrected images are superior and look reasonable. As for each acquired k-space core, a full image can be reconstructed, our method yields a time-resolved 4D image. The quantitative validation of this method using an anthropomorphic phantom with repeatable respiratory motion is planned for the future.

Acknowledgments
This work was supported by the Swiss National Science Foundation, SNSF (project number: 320030 163330/1).

References
[1] Stemkens et al. (2018), PMB(63); [2] Krieger et al. (2019), ESTRO38; [3] Jud et al. (2018), MICCAI18

Speaker: Dr Christoph JUD (University of Basel)

11:15 Ultrasound-based Lung Motion Modelling

Introduction
Respiratory motion poses great challenges in pencil beam scanned (PBS) proton therapy of mobile targets. In a recent study, we presented the potential of tumour tracking using a patient-specific motion model on simulated 4DCT(MRI) data sets [1]. A statistical motion model was used to estimate dense lung motion information from 2D abdominal ultrasound (US). While this study was based on Gaussian process (GP) regression [2], an alternative approach using cubic polynomial regression could be used likewise [3]. Compared to GP models, polynomial regression models have the advantage of reduced computational complexity since they can be addressed with ordinary least squares analysis. In this work, we aim to compare the two US-based motion models using the 4DCT(MRI) data sets presented in [1].

Materials and Methods
Simultaneous acquisitions of abdominal US imaging and time-resolved 4DMRI [4] were performed on two healthy volunteers on a 1:5T clinical scanner (MAGNETOM Aera, Siemens Healthineers, Erlangen, Germany). Continuous acquisition during 10 min of free breathing resulted in approximately 700 distinct respiratory states. Deformable image registration was applied to compute the deformation vector field (DVF) with respect to a reference exhalation volume for each respiratory state. These DVFs were then used to animate full-exhale CTs of two lung cancer patients, resulting in four synthetic 4DCT(MRI) data sets [1].
Low-dimensional respiratory surrogate signals were extracted from the 2D US image series using principal component analysis (PCA). For the GP regression model, PCA was also performed on the lung deformation field. The data sets were split into a training and test set, comprising 500--650 and 50--73 US/MR image pairs, respectively. Both, the GP regression model and the cubic polynomial regression model were evaluated on the same data sets and compared below.

Results
We defined the prediction error as the magnitude of the difference between the reference and the predicted DVFs. Figure 1 shows the distribution of the mean and the 95th percentile prediction error. It can be seen that the GP model outperforms the cubic polynomial model in all data sets.

Conclusions
This analysis suggests that the GP model is more accurate than the cubic polynomial model which comes however at the cost of increased computational complexity. In the context of radiotherapy, uncertainty quantification using GP models could be used to monitor motion prediction confidence and if necessary to interrupt the treatment. In a future work we will investigate the performance of the GP model for beam adaption in PBS proton therapy of lung tumours.

Acknowledgements
This work was supported by the Swiss National Science Foundation, SNSF (project number: 320030 163330/1).

References
[1] Krieger, Giger et al. (2019), ICCR 19; [2] Rasmussen and Williams (2006), MIT Press; [3] Giger et al. (2019), accepted for presentation at PRIME-MICCAI19; [4] von Siebenthal et al. (2007), PMB(52)

Speaker: Ms Alina GIGER (Center for medical Image Analysis & Navigation, University of Basel, Allschwil, Switzerland; Department of Biomedical Engineering, University of Basel, Allschwil, Switzerland)

11:30 → 12:15 Invited Speaker Prof. Rob Coppes: The use of tissue resembling organoids to predict prevent and repair radiation responses

University Medical Center and University of Groningen, Groningen, The Netherlands

12:15 → 13:30 Lunch

13:30 → 14:45 Session II: Miscellanea - Chair: Sheeba Thengumpallil

Clinique des Grangettes

13:30 Impact of electromagnetic stray radiation in the hyperthermia treatment room

Introduction
Hyperthermia to temperatures of 41-43°C has been shown to be a valuable sensitizer for radiation- and chemo-therapy in cancer treatment. During hyperthermia treatment (HT) the presence of health staff and accompanying persons in the treatment room is desirable for patient comfort. In this study the electromagnetic stray radiation in the treatment room with respect to limits for non-ionizing radiation was investigated.

Material and Methods
The strength of the electromagnetic field inside the unoccupied HT room was measured with a calibrated power density meter (RAHAM Model 4C, General Microwave Corporation Amityville (NY), USA). Measurements were made on a 2-dimensional grid with measurement point 20cm apart. During the measurements the HT machine (BSD2000-3D with SigmaEye applicator, Pyrexar Medical, Salt Lake City, USA) was running with a dummy load of saline water. Power at the generator output was 1000W maximum, distributed through the feeding network to 24 antennas of the SigmaEye applicator. The results of the electrical field strength measurements were evaluated according to the ICNIRP guidelines (1).

Results
A clear linear relation between output power at the generator and measured power density at a single measurement point was found. Inside the HT room electrical field strength values between 17.4 V/m and 137.3 V/m were detected. According to the ICNIRP guidelines (1) areas with electrical field strength below and above the limit for occupational exposure of 61 V/m were detected and marked on the floorplan of the HT room for a typical patient treatment generator output power of 800W (Figure1).

Conclusions
In large areas of the empty HT room the limits for occupational exposure are exceeded. Specific areas around the treatment table were identified as safe. Since the presence of persons and objects inside the HT room strongly influences the measured field strength further measurements should investigate the actual field strengths during patient treatment and indicate measures to avoid high E-fields.

References
(1) ICNIRP Guidelines for Limiting Exposure to Time-Varying Electric, Magnetic, and Electromagnetic Fields (up to 300 GHz). In: Health Physics 74 (4): 494–522; 1998

Speakers: Mr Dietmar MARDER (Kantonsspital Aarau, RadioOnkologieZentrum KSA-KSB), Mr Denis OVCARI (Kantonsspital Aarau, RadioOnkologieZentrum KSA-KSB)

13:45 Clinical impact of metallic surgical clips in superficial radiative hyperthermia

Introduction
Mild hyperthermia (HT) aims to heat tumour tissue to 40-44°C for 1 h and has proved to increase the therapeutic effect of radiotherapy (RT) with, among others, excellent clinical results for local recurrence of breast cancer. Two patients in our department with breast cancer recurrences were good candidates for combined RT-HT, but had surgical clips in the region to treat. Metallic implants are a contra-indication for radiative HT because of a risk of localized hot spots near the metal, with possible overheating and damaging of surrounding tissue. The sparse literature on the subject prompted us to investigate the difference in heating pattern caused by the clips, using a specific HT phantom.
Material and Methods
Patient 1 (P1) had eight 6-mm long titanium clips (type T1), implanted in the breast, at depth ranging from 2.5 to 3.5 cm. P2 had a single 7.1-mm long clip of type O’Twist axillar (type T2), made of a nickel-titanium alloy, implanted at 3 cm depth in an axillar node. In both cases, the clips were located in the central region of the HT field. Superficial HT in our department is delivered with the microwave (MW) Alba HT system operating at 434 MHz. The machine is equipped with different size antennas. The antennas we used in this study were the ones used for the treatment of P2 and P1, i.e. alpha and beta respectively (with effective thermal field size of 4x16 and 8x12 cm2, respectively, and thermal penetration depth of 3 cm). Measurements of the difference in temperature rise (TR) caused by the clips presence compared to a setting without clips were performed at 1 cm depth (depth of the max of TR) in slabs of a home-made Agar muscle-equivalent phantom. We used 12 T1 clips positioned in different configurations (clips aligned, clustered, etc.) and a single T2 clip. Temperature measurements were performed real-time (i.e. while heating) with thermocouples (TC) positioned next to the clips, and after heating intervals of 6 and 20 minutes with a calibrated thermal camera. The maximum power (100 W) was used for all measurements.
Results
TC provided limited point information and interpretation of results was complicated by the intrinsic inhomogeneity of TR across the antenna. In contrast, the thermal camera gave 2D thermal measurements with high resolution, but required the removal of the antenna from the phantom. TC measurements were sufficient to exclude a significant MW absorption (leading to a ΔTR>2°C) for any clips types and configurations, while camera results allowed to exclude the occurrence of local hotspots (ΔTR>0.5°C). Based on these results, we evaluated the risk of serious adverse effects for patients to be negligible. P1 and P2 were treated with 5 and 3 HT sessions, respectively, with no noticeable effects related to the clips.
Conclusion
Results suggest that superficial radiative HT is safe in patients with implanted surgical clips of the tested types. Two patients with implanted clips benefited from HT in our department.

Speaker: Dr Maud JACCARD (Geneva University Hospital)

14:00 Evaluation of a MR compatible phantom for patient specific QA

Introduction:
Most phantoms for patient specific Quality Assurance (QA) are not MR compatible and not MR safe. The Delat4+MR phantom was redesigned to be used in combination with Magnetic Resonance Linear Accelerators (MR-Linac). Here we report extended testing of this phantom and our early clinical experience using the phantom for patient QA.

Materials and Methods:
Since April 2019, a 0.345T MR-Linac (ViewRay, Mountainview, SN228) is clinically operational in the Radiation Oncology department at University Hospital of Zürich. For patient specific QA the Delta4 Phantom+MR (ScandiDos, Sweden) has been tested. With this phantom and a MR compatible Farmer Ionization Chamber (PTW Freiburg, Type TW30013), four tests have been performed. 1. Daily output of the MR-Linac: The output of the machine was measured daily with the Delta4+MR phantom and the chamber in a solid water block. 2. Angular dependency of the phantom: Equally distributed fixed beams with different gantry angles were irradiated on the Delta4+MR phantom and compared with free air measurements of the Farmer Chamber. For the Delta4+MR, the lower beams (gantry angles 135° to 225°) were absorbed additionally by the couch and the fixed compartment of the couch, whereas for the Farmer chamber only the latter was the case. 3. Field size dependence of the phantom: Square fields from 4x4cm2 to 20x20cm2 were irradiated. Due to the plus arrangement of the 2D arrays of the Delta4+MR phantom, the field size dependency measurements were conducted at a gantry angle of 315°. 4. Patient specific QA: The Delta4+MR phantom was used to asses several patient plans. All the measurements were evaluated with the gamma evaluation method in the Delta4 software (by ScandiDos, Sweden).

Results:
The daily output changes measured with the phantom were within 2%(0.9%) and in good agreement with the chamber measurements(0.6%). The angular dependency of the phantom was small (average gamma passing rates 98% for 3%/3mm and 94% for 2%/2mm criteria) except for the gantry angles 0°,180°, 90 and 270°. This is probably due to the ‘plus’ detector arrangement of the phantom. Excluding these four angles, the gamma agreement increased to 99.2% for 3%/3mm criteria. The average deviation for angular dependence between Delta4 Phantom+MR and Farmer Chamber was 0.5%. The phantom’s gamma passing rate for a field size of 4x4cm2 was 100% for 3%/3mm and 92.8% for 2%/2mm criteria and 99.9% for 3%/3mm and 94.9% 2%/2mm for the field size of 20x20cm2. The disagreement was similar for all field sizes. Out of the first measured 36 patient plans, only one plan failed the gamma passing criteria of >95% (3%/3mm), the average gamma passing rate was 99.7% for 3%/3mm and 98.2% for 2%/2mm criteria.

Conclusion:
It was shown that the MR compatible phantom is stable over time and shows only a small directional and field size dependence. The Delta4 Phantom+MR can be used for patient specific QA without any major concerns.

Speaker: Ms Yasaman SAFARKHANLO (University of Zurich and UniversitätsSpital Zürich)

14:15 Characterization of 3D spatial integrity on a 0.35T MR-LINAC

Introduction

The MRIdian Linac (ViewRay, Ohio, USA) is a MR-guided radiotherapy system which combines a low field MR scanner (0.35T) with a radiotherapy gantry (6MV linac) located in the magnet split bore. The superior high soft tissue contrast of MRI compared to computed tomography is expected to improve tumor delineation as well as patient setup. However, the geometric accuracy that is crucial in radiation therapy to place the dose at the intended position, is limited by several hardware factors (magnetic field inhomogeneities, gradient non-linearity) as well as by sample related effects (susceptibility artefacts, chemical shift). This abstract aims at characterizing the 3D spatial integrity of the MR images at different gantry angles and for different offsets positions in the transverse plane.

Materials and Methods

We acquired images of the Large Field MRI Distortion Phantom (CIRS, Norfolk, USA) with the clinically bSSFP (balanced steady state free precession) sequence modified to cover a large field of view (FOV) of 32.4x35x35.2cm3 with a resolution of 1.048x1.048x1mm3.This phantom consists of a 3D grid (3mm rodes spaced by 20mm, submillimeter precision) distributed in a cylinder like structure (diameter 330mm, length 300mm) filled with water. Over the whole volume the grid vertices define 1885 control points. The phantom was imaged at four gantry angles (0°, 90°, 180° and 270°) and at four different shifts in the transverse plane. We implemented a program in Matlab (R2019a, The MathWorks, USA) to automatically detect the position of the vertices in the acquired volume inspired by the methodology presented by Jafar et al. [1]. This software relies on a pattern matching approach combined with a connected component algorithm. The set of vertices representing the phantom was rigidly registered to the theoretical locations of the vertices to correct for small rotations and translations in the phantom setup. Spatial integrity is assessed by measuring the Euclidian distance of the detected vertices to their theoretical locations. Manual post-processing was done in order to remove markers located in corrupted image region (especially due to bSSFP banding artifact).

Results

The passing rate specified by the vendor are 1mm and 2mm within a 100mm and 175mm radius from the isocenter, respectively. The maximum displacement distances measured for each configuration (table 1) respect these constraints except for the 0° gantry angle configuration. However, the average distortion for this setup is 0.44mm. Ginn et al. [2] reports similar results for the Cobalt-60 MRIdian model. There was no change in distortion when the phantom was shifted +/-40mm in the transverse plane.

Conclusions

The specification defined by the vendor are met and could be assessed simultaneously in three dimensions for different clinically relevant setups. Patient positioning can be performed off-axis without increased distortions.

References:
1. Jafar, M., et al., Assessment of Geometric Distortion in Six Clinical Scanners Using a 3D-Printed Grid Phantom. 2017. 3(3): p. 28.
2. Ginn, J.S., et al., Characterization of spatial distortion in a 0.35 T MRI-guided radiotherapy system. Physics in Medicine & Biology, 2017. 62(11): p. 4525.

Speaker: Dr Bertrand POUYMAYOU (Radiation Oncology Department, Universitätsspital Zürich, Switzerland)

14:30 Tomotherapy implementation and integration in RayStation: one year-experience.

Introduction: To report about the Tomotherapy implementation and integration in RayStation Treatment Planning System (TPS) version 8a. This new mixed system involved the planning with Raystation and then the export of treatment plans to iDMS, a single database provided by Accuray common for both machines.
Materials and Methods: Two Tomotherapy units were commissioned in November and December 2018, respectively. Machine models imported in RayStation were expected to comply with the gold standard models unless minor adjustements of the calculated output factors for each field width. Machine models were dosimetrically validated via in-phantom measurements of plans designed for each field width and for both techniques, TomoHelical and Tomo Direct, by means of A1SL ionisation chamber. In addition, a variety of clinical Tomotherapy-based plans designed with RayStation was checked using the PTW Octavius system. Furthermore, any difference observed with respect to the previous system (i.e. Accuray only, both TPS and database) in its clinical use was also registered and herein reported.
Results: Adjustments in the output factor were observed < 1% ranging between 1 and 1.01 for one machine and 0.995 and 1.006 for the second one. With such parameters set in the machine models, difference between calculated and measured dose was observed less than 1.5% as maximum and less than 0.5% in average. Clinical plans were successfully delivered having a gamma pass index > 97% with 3%/3mm criteria. Several differences were observed when comparing the clinical use with the previous clinical implementation: 1) Machine maintenance could cause a discrepancy between the model existing in the TPS and the one in iDMS avoiding treatment unless remodelling the machine in RayStation. 2) After a number of times (> 80), plans QA could not be longer exported to iDMS unless changing their DICOM properties. 3) Treatment plans having low (0.5 Gy) prescription doses must be planned using a machine model encompassing the full range of leaf-open-time (LOT) and not only limited to 60ms as pre-configured in RayStation. 4) Treatment plans requiring all the leaves of the MLC working simultaneously required more additional planning time to fulfil the mechanical limits of the machine. 5) Patients transfer between machines was not available in this current version of the TPS.
Conclusion: The commissioning of both machines in RayStation was straightforward with no complications. On the other hand, the integration between the version 8a of the TPS and iDMS suffered from several pitfalls, which could have a significant impact on the clinical routine.

Speaker: Michele ZEVERINO (Institute of Radiation Physics, Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland)

14:45 → 15:30 Session III: Dosimetry I - Chair: Dr. Sheeba Thengumpallil

Clinique des Grangettes

14:45 First Voxel-wise Prediction of Post-therapy Dosimetry for 177Lu-PSMA I&T Therapy

Introduction:
PSMA-directed radioligand therapy (RLT) has become one of the effective treatment options for metastatic castration-resistant prostate cancer (mCRPC). However, individual treatment planning is still not feasible as it is for the external beam radiotherapy. Our group has presented the first organ-based research in the prediction of post-therapy dosimetry in SSRMP 2018. However, an organ-based approach is unable to reveal the heterogeneity of dose distribution and therefore is not sufficient for the realization of treatment planning. In this study, we propose the first approach for voxel-wise prediction of post-therapy dosimetry via generative adversarial networks (GANs) from pre-therapy positron emission tomography (PET) images.
Materials and Methods:
30 patients with mCRPC treated with 177Lu-PSMA I&T RLT were retrospectively included in this study. Only those cycles with 68Ga-PSMA-11 PET/CT directly before the treatment and at least 3 post-therapeutic SPECT/CT dosimetry imaging were selected. Totally 48 treatment cycles were considered for this proof-of-concept study. 3D RLT Dose GANs were developed with a 3D U-net generator and a convolutional neural network (CNN) based discriminator. An advanced dual-input-model was designed to incorporate both information from PET and CT, for the purpose of anatomical coregistration. Both voxel-wise content loss alongside image-wise loss were taken into account for better synthesis performance. K-fold cross validation was applied to verify the trained network.
Results:
The proposed 3D RLT Dose GANs achieved the voxel-wise mean absolute percentage error (MAPE) of 17.56%±5.42%. The dual-input-model was able to synthesize dose maps with comparable accuracy while preserving anatomical consistency, which achieved a MAPE of 18.94%±5.65%.
Conclusions:
Our experimental results demonstrate the capability of artificial intelligence to estimate voxel-wise post-therapy dosimetry both qualitatively and quantitatively, may provide a practical solution to improve the dosimetry-guided treatment planning for RLT.

Speaker: Song XUE (Dept. Nuclear Medicine, University of Bern, Switzerland)

15:00 EPR dosimetry in the radiotherapy dose range using a benchtop spectrometer: Dose uncertainties for alanine and lithium formate dosimeters.

Introduction
In electron paramagnetic resonance (EPR) dosimetry, water equivalent dosimeter materials such as alanine (Al) or, more recently, lithium formate monohydrate (LFM) featuring low fading rates (a few percent per year) and a linear dose response are typically used. The dependency of the dosimeter response on beam energy, dose rate and beam angle is usually negligible for therapeutic beams offering high potential for applications in radiotherapy. Moreover, EPR dosimetry provides a non-destructive readout allowing repetitive measurements. Despite its favorable characteristics, EPR dosimetry is not widespread in the clinics due to the reported sensitivity limitations at the therapeutic dose level and the associated efforts. In the present work, we compared commercial Al pellets with self-manufactured LFM pellet dosimeters regarding the dose uncertainty in the dose range from 1 to 70 Gy by using a 'practical' EPR dosimetry system.

Materials and Methods
The Al pellets were purchased from Aerial, whereas the LFM pellets were pressed in-house. All pellets had a diameter of 4 mm and a height of 2 mm (Al) or 4 mm (LFM). The pellets were irradiated to doses of 1, 5, 20, 50 or 70 Gy by a clinical 6 MV photon beam. In total, 25 pellets per material (5 per dose value) were examined. For each pellet, five independent EPR measurements were performed on a benchtop EPR spectrometer (MiniScope MS 5000) within five weeks following irradiation. The measurement time of a single readout was restricted to 10 min per pellet. Dose values were reconstructed from EPR signal amplitudes using an in-house developed spectral fitting procedure.

Results
In terms of dose uncertainty, the self-made LFM pellets are superior to the commercially available Al pellets, mainly due to the higher EPR signal intensity resulting from the increased pellet mass and the narrower EPR spectrum. The relative dose uncertainties (1σ) for a single readout at doses ≥ 5 Gy are below 2.8 % (Al) and 1.0 % (LFM) but increase to 12.3 % (Al) and 2.6 % (LFM) at 1 Gy. The uncertainties at 1 Gy decrease to 2.6 % (Al) and 0.8 % (LFM) when five independent readouts are averaged.

Conclusions
In the case of the LFM pellets, the EPR dosimetry system shows a high level of precision down to 1 Gy being suitable for routine QA applications in radiotherapy (e.g. in-vivo or small field dosimetry). The uncertainties can be further decreased by averaging several independent measurements. This rather time-consuming procedure is especially advisable when using the commercial Al pellets.

Speaker: Mr Sebastian HÖFEL (Department of Chemistry and Konstanz Research School Chemical Biology, University of Konstanz, Germany)

15:15 Dosimetric procedures for FLASH-RT biological experiments

Introduction
Recently, some of the established concepts governing the effects of radiation on healthy tissues were questioned by a new treatment modality using ultra-high dose-rates called FLASH radiotherapy (FLASH-RT). To this day, no metrological traceability for FLASH-RT dosimetry exists and no monitoring instruments can determine on-line the delivered dose during biological experiments. The aim of the study was 1) to conduct a redundant evaluation of the dose with different measurement means, as a surrogate for metrological traceability and 2) to implement a dosimetric procedure for biological experiments.
Materials and Methods
The experiments were conducted on the Oriatron eRT6 (PMB-Alcen, France), a prototype high dose-per-pulse linear accelerator delivering a 6 MeV pulsed electron beam with dose-rates ranging from conventional (a few Gy/min) to ultra-high (>1000 Gy/s). To ensure minimal uncertainties in the delivered dose, a two-phase procedure was developed. First, in order to achieve traceable, repeatable and stable irradiations, a redundant dosimetry with alanine, Thermo-Luminescent Dosimeters (TLD) and radiochromic films was conducted to investigate absolute dosimetry for FLASH-RT. Then, for biological experiments, a dosimetric procedure was implemented prior to irradiation to ensure the delivered dose. Three typical biological setups are presented together with the beam characteristic and dosimetry.
Results
Alanine, TLDs and films are suitable dosimeters for absolute dosimetry for irradiations with dose-rates between 0.078 Gy/s and 1050 Gy/s (dose agreement of 3% between them). When applying our dosimetric procedure prior to irradiation, the maximum deviation between the delivered and prescribed dose was less than 3%. Dose deviations up to 15% of the prescribed dose could be achieved in the absence of setup corrections (i.e. without dosimetric procedures).
Conclusion
Thanks to the validation of several dosimetric means, we were able to develop procedures to accurately irradiate biological models. An agreement of 3% between the delivered and prescribed dose was achieved.

Speaker: Patrik JORGE (IRA)

15:30 → 16:00 Coffee break

16:00 → 16:45 Invited Speaker Prof. Gunther Dissertori: From Particle Physics to (Bio-)Medical Imaging: A personal travelogue

Institute for Particle Physics and Astrophysics, ETH Zürich, Zürich, Switzerland

16:45 → 17:45 General Assembly

18:30 → 20:00 Apéro starting from 6pm

20:00 → 23:00 Social Event

Friday 22 November

09:00 → 09:45 Session IV: Dosimetry II - Chair: Dr. Daniel Vetterli

Radio-Onkologiezentrum Biel / Inselspital

09:00 Calibration and verification of dosimeters for high energy electron and photon radiation at METAS 22 MeV accelerator

1 Introduction

Therapy dosimeters for high-energy photon and electron radiation have to be verified every four years by METAS. The verification consists of a general test of the equipment, the determination of the reference value of the check-source and, most essentially, the actual calibration of the dosimeter (in following device under test, DuT).
METAS has two different calibration methods available:

1. Using Co-60 radiation and applying pre-determined constants to calculate the calibration factors (kQ model) for the various beam qualities used in hospitals.
2. Direct comparison of DuT with a secondary standard in high energy electron or photon field of the 22 MeV electron accelerator. Since the accelerator has more than eight different photon and electron beam qualities available, DuTs can be calibrated at beam qualities very similar to those used in hospitals.

In the past, most calibrations were performed with Co-60 for efficiency reasons. However, recent progress in the accelerator performance now allows for very efficient calibration in accelerator beam. Moreover, a new measuring station was developed which will be presented here.

2 Materials and Methods

METAS operates a flexible 22 MeV Microtron-type electron accelerator for calibration and research purpose. It is equipped with a conventional treatment head. It serves as a source for clinical electron beams (R50 = 1.75 to 8.54g/cm2) and photon beams (TPR20,10 = 0.639 to 0.789), respectively.
The comparison between the secondary standard and the customer's device (DuT) is done indirectly via a set of two monitor chambers: the calibration procedure sequentially irradiates either the secondary standard or one DuT, whereas the two monitor chambers are used for the normalization. A new software that also monitors environmental parameters and regularly checks the accelerator status manages this sequence in an automated way. In addition, the software is connected to a database containing all the relevant measurement data and parameters. It also keeps track of the measurement devices and their calibration certificates.

3 Results

The new calibration station at METAS' accelerator substantially simplifies the calibration process and it enables ionization chambers to be calibrated as efficiently as in Co-60. The new measuring station was validated by comparing several secondary standards. Good agreement was found to the results obtained during the last primary campaign.

4 Conclusions

The new calibration station enables an efficient service for the calibration of therapy dosimeters directly in high-energy electron and photon radiation but with reduced uncertainty compared to the method in Co-60. It is foreseen that this service will be offered to the verification customers in future.

Speaker: Dr Peter PEIER (METAS)

09:15 Chemical radiation dosimetry in magnetic fields: Characterization of a Fricke-type chemical detector in 6 MV photon beams and magnetic fields up to 1.42 T

Introduction
The ideal detector for MR guided radiotherapy (MRgRT) applications is not influenced by the presence of a magnetic field during the irradiation. The Fricke-type detector is a chemical dosimetry method in which ferrous iron (Fe$^{2+}$) is oxidized to ferric iron (Fe$^{3+}$) upon irradiation. The increase of the Fe$^{3+}$ concentration is proportional to the absorbed dose and it is not expected that the chemical reactions are greatly influenced by the magnetic field.

Materials and Methods
The Fricke detector consists of a perfluoroakoxy alkane (PFA) sample cup filled to the top with an acidic ferrous iron solution. The detector readout was performed with a UV spectrophotometer (Varian Cary 6000i) by measuring the absorbance change of the solution upon irradiation and calculating the change of the Fe$^{3+}$ concentration (Δc) using the Lambert-Beer law. The irradiations were performed in a 6 MV photon beam of a linear accelerator (Elekta Precise Treatment System) using a dose rate of 2 – 3 Gy/min and a pulse repetition frequency of 400 Hz. A magnetic field was generated in between the pole shoes of a constant-current driven electromagnet (Bruker ER0173W), which was homogeneous within 1 μT over 1 cm$^3$ region up to 1.42 T.

Results
The linearity of the Fricke response curves between 6 Gy and 100 Gy is not influenced by the applied magnetic fields (0 T – 1.42 T) during the irradiation with a 6 MV photon beam. The magnetic field correction factors k_B for the Fricke detector for magnetic field strengths of 0.35 T and 1.42 T are 0.995±0.005 and 0.998±0.005, respectively.

Conclusion
The response of the Fricke detector is not significantly influenced by the presence of a magnetic field during the irradiation and the correction for the applied magnetic field is <0.5%. This correction is rather small in comparison to ionization chamber type detectors, which have corrections up to several percent dependent on the mutual orientation of the detector and the magnetic field.

Speaker: Maria TRACHSEL (Federal Office of Metrology METAS)

09:30 Quality Assurance of a new accelerator with an integrated MR system

Introduction
The introduction of the MR-Linac technology poses new challenges on the definition of a suitable machine Quality Assurance (QA) program. In this work, we present our current QA program for the MRIdian MR-Linac (Viewray) and the preliminary results on machine performance.

Materials and Methods
The acceptance of the MRIdian was performed in December 2018. Since then, we acquired data using MR-compatible ion chambers, StarCheck MR (PTW) and Gafchromic® films. In order to check the correlation between MR and RT isocenter, we used the DailyQA phantom (Viewray). ACR, NEMA and distortion measurements were performed to check the MRI quality and stability. We adapted the SGSMP recommendations for Linac QA to the characteristics of this new machine and devised a QA program to check the Linac and MRI performances.

Results
The output stability is measured daily with an ion chamber. Over a period of six months, the stability was better than 1%. The symmetry is measured weekly using the Starcheck MR and it was stable over the same period of time; being 1.8% and 0.7%, in the LR and GT directions. MLC positional accuracy is evaluated daily at gantry zero and monthly at gantry 90° and 270°, using the picket fence test, and it showed a maximum deviation of 0.5±0.1 mm. The size of the treatment isocenter is evaluated weekly with star shots, and it has a diameter of around 0.7±0.1 mm in the axial plane. The deviation between MRI and radiation (RT) isocenter was evaluated using the Viewray daily phantom and it was 0.7±0.3 mm. ACR, NEMA and distortion measurements were performed weekly for the first four months. At present only the ACR is performed weekly and the other checks monthly. Additionally, a quick MRI distortion check is performed daily.

Conclusions
The SGSMP recommendations covered satisfactorly the QA requirements for the MR-Linac technology with the exception of the MRI QA. Viewray MRIdian proved to be stable over a six-month period. The size of the treatment isocenter, the MLC positional accuracy and the agreement between MR and RT isocenters are on the submillimeter range.

Speaker: Dr Mariangela ZAMBURLINI (Klinik für Radio-Onkologie, UniversitätsSpital Zürich)

09:45 → 10:45 Session V: Radiology - Chair: Dr. Daniel Vetterli

Radio-Onkologiezentrum Biel / Inselspital

09:45 Online tool for fast and accurate assessment of radiation dose received by the fetus in CT

Introduction
Computed tomography (CT) is sometimes needed during pregnancy. In this case, assessment of radiation dose received by the fetus is required. Existing methods are either cumbersome or limited in their accuracy. The aim of this study is to develop and validate computational algorithm for fetal radiation dose assessment, which could be used in clinical routine.

Materials and methods
First, commercially available Monte Carlo code was modified in order to simulate dose distribution from a vendor-independent (generic) CT system. Then, this code was validated against the CTDI measurements performed on CT systems of two different manufacturers (Siemens and GE). A set of Monte Carlo (MC) simulations with various exposure parameters was performed on computational phantoms representing pregnant patients at various gestational stages. The normalized fetal dose values from these MC simulations were recorded. These normalized dose values were used for the computational algorithm enables fetal dose assessment from CT examination of various body region, at different exposure settings.
In the second stage of the study, the accuracy of the proposed algorithm was validated against detailed MC simulations performed on CT data from 29 real pregnant patients underwent clinically indicated abdominal CTs. After the validation, a user-friendly online tool was developed.

Results
The median fetal dose from abdominal CT calculated for the real pregnant patients was 2.7mGy. The relative error of the dose values calculated by the online tool was 11%, on average. The highest error of about 40% was found in patients with additional hardware (i.e. fixation device) resulted in the increased tube current (mA) applied by the system and thus higher fetal dose estimated by the program.
The online tool (www.fetaldose.org) enables vendor-independent calculations of fetal doses at various gestational ages from CT examination of any type.

Conclusion
The tool provides fast and reliable evaluation of radiation dose, received by the fetus from CT examination of the mother. This tool requires the input of only a few parameter and can be used by clinicians in their routine.

Speaker: Dr Natalia SALTYBAEVA (University Hospital Zurich)

10:00 Development of semi-classical Monte Carlo techniques for the simulation of clinical grating interferometry breast CT modules

Introduction

Grating interferometry (GI) breast CT (GI-BCT) aims to improve breast cancer diagnosis utilizing the enhanced soft tissue visibility by combining absorption, phase and dark-field contrast and acquiring three-dimensional data of the uncompressed breast. Due to interference and scattering phenomena in an X-ray GI and the use of incoherent X-ray sources there is no established simulation tool for accurate estimations of image quality, scattering contributions, beam spectra and patient dose, which are important quantities in the design process of clinical GI-BCT modules. Previous approaches to X-ray GI simulations, as for instance Monte Carlo (MC) algorithms implementing Huygens principle, can adequately simulate the required interference phenomena, but mainly struggle with consistent description of all involved processes, modeling incoherent sources, scattering and suffer from long computation times.

Materials and Methods

In this work a MC algorithm was developed based on quantum mechanical principles including classical approximations and implemented as an extension library of the well-established MC particle transport code EGSnrc. As a test for the capability of the developed MC algorithm to simulate interference and attenuation effects, a simplified setup with a sample in front of a silicon pi-phase grating is simulated and the absorption and differential phase images are retrieved. The sample consists of two polystyrene spheres of different sizes, one of which with a silicon core and is illuminated by a monochromatic plane wave in a 1 by 1 millimeter field of view.

Results

The retrieved absorption and differential phase projections have the expected features of the imaging method. The silicon core is clearly visible in both absorption and phase contrast, while the weakly absorbing polystyrene spheres are only visible in the differential phase projection. A performance comparison to an in-house reimplementation of a previously published Huygens principle MC algorithm, in a much smaller simulation volume, shows that the presented algorithm is roughly 2.5 to 25 times faster depending on transport parameter choices.

Conclusion

The newly developed algorithm relies on a consistent physics model of all involved processes and shows desired behavior in academic simulation setups. The quantum mechanical foundation of the algorithm allows to decrease simulation time and the extension to incoherent X-ray sources. Future extensions of the framework will include methods for dose estimation and performance relevant simplifications for the simulation of micrometer sized scatterers required for dark-field simulations.

Speaker: Mr Stefan TESSARINI (Institute for Biomedical Engineering, University and ETH Zürich, 8092 Zürich, Switzerland; Swiss Light Source, Paul Scherrer Institut, 5232 Villigen, Switzerland)

10:15 Longitudinal radiomics using CBCT for prognosis in non-small cell lung cancer

Introduction
The prognostic value of radiomics for non-small cell lung cancer (NSCLC) patients has been investigated for images acquired prior to treatment, but no prognostic model has been developed that includes the change of radiomic features during treatment. Therefore, the aim of this study was to investigate the potential added prognostic value of a longitudinal radiomics approach using cone-beam computed tomography (CBCT) for NSCLC patients.

Materials and methods
This retrospective study includes a training dataset of 141 stage I-IV NSCLC patients and three external validation datasets of 94, 61 and 41 patients, all treated with curative intended (chemo)radiotherapy. The change of radiomic features extracted from CBCT images was summarized as the slope of a linear regression. The CBCT slope-features and CT-extracted features were used as input for a Cox proportional hazards model. Moreover, prognostic performance of clinical parameters was investigated for overall survival and locoregional recurrence. Model performances were assessed using Kaplan-Meier curves and c-index.

Results
The radiomics model (Figure 1, model 1.1) contained only CT-derived features and reached a c-index of 0.63 for overall survival and could be validated on the first validation dataset. No model for locoregional recurrence (model 1.2) could be developed that validated on the validation datasets. The clinical parameters model could not be validated for either overall survival (model 3.1) or locoregional recurrence (model 3.2).

Conclusions
In this study we could not confirm our hypothesis that longitudinal CBCT-extracted radiomic features contribute to improved prognostic information. Moreover, performance of baseline radiomic features or clinical parameters was poor, probably affected by heterogeneity within and between datasets.

Speaker: Dr Janita VAN TIMMEREN (The D-Lab: Decision Support for Precision Medicine, GROW - School for Oncology and Developmental Biology, Maastricht University Medical Center, Maastricht, the Netherlands)

10:30 Evaluation, comparison, and optimization of the effects of manual versus software-automated protocols on radiation dose and image quality in simulated paediatric chest computed tomography

Background: The aim of this study was to compare the effects of switching from automated to manual acquisition parameters on image quality and radiation dose in several simulated paediatric chest computed tomography (CT) scans. The comparison was made using a CT phantom in order to obtain the lowest possible dose–length product value while maintaining an adequate image quality. The scans were performed by manually reducing the dose below the lowest dose value proposed by automated software prior to the examination.

Methods: An anthropomorphic phantom underwent simulated paediatric chest CT scans using both automated and manual approaches guided by a radiologist for parameter optimization. Different CT acquisition protocols were used, varying kV, mAs, pitch, and adopting iterative reconstruction (IR). The subjective and objective image qualities were assessed by, respectively, radiologists and software. Specific CT dose indices were collected.

Results: CT dose indices were significantly lower adopting a manual approach. Through CT acquisitions, linearity and resolution were quite constant, whereas image noise and uniformity varied between scans, as observed by radiologists using a visual grading analysis. IR was associated with a further dose reduction.

Conclusion: Simulated paediatric chest CT studies performed with manual acquisition settings resulted in important dose reduction when compared to values generated with automated protocols.

Speakers: Dr Gianluca ARGENTIERI (Ente Ospedaliero Cantonale – IIMSI - Servizio di Radiologia del Sottoceneri - Ospedale Civico Lugano), Dr Luca BELLESI (EOC)

10:45 → 11:15 Coffee break

11:15 → 12:00 Invited Speaker Prof. Thomas Bortfeld: Democratizing proton therapy

Harvard Medical School and Massachusetts General Hospital, Boston, USA

12:00 → 13:15 Lunch

13:15 → 14:30 Session VI: Protons - Chair: Francesca Belosi

Center for Proton Therapy at Paul Scherrer Institut

13:15 Combined proton-photon treatments: How can limited proton slots be optimally distributed over a patient cohort?

Introduction: Although rapidly growing, proton therapy is a limited resource, which is not available to all the patients who may benefit from it. In this study, we investigate if combined proton-photon treatments, in which some fractions are delivered with protons and the rest with photons, improve on single-modality treatments. Combined treatment can be motivated by the consideration that, on the convex part of the NTCP curve, the first proton fractions are the most beneficial. We assume a situation of limited proton slot availability and develop methods to distribute those limited slots over a patient cohort optimally in order to optimize the benefit of proton therapy at a population level.

Material and Methods: We consider a cohort of 45 head and neck cancer patients for which IMRT and IMPT plans were previously created [1]. NTCP models for relevant side effects (e.g. xerostomia) were used to calculate the NTCP values for all the plans. We investigate a 30 fraction simultaneous integrated boost (SIB) scheme (1.8 Gy to the PTV, 2.3 Gy to the GTV) and a sequential boost (SEQ) scheme with a 25 fraction base plan (2 Gy to the PTV) and a 10 fraction boost plan (2 Gy to the GTV). Under the assumption that, due to limited resources, only a small percentage of the total number of fractions can be delivered with protons, an integer programming algorithm was applied to determine the optimal number of proton fractions per patient that minimizes the total number of expected complications over the patient cohort.

Results: Figure 1a shows the NTCP values for xerostomia in the IMRT and IMPT plans for the SIB scheme for all patients. Figures 1b shows the optimal allocation of proton slots for the situation where 20$\%$ of all fractions are delivered with protons. The patients with the highest $\Delta$NTCP value (IMRT-IMPT) receive the largest number of proton fractions. The average xerostomia NTCP value from all 45 patients for the SIB scheme for combined treatment equals $13.0 \%$. For the single-modality treatment, where the $20\%$ of patients with the highest $\Delta$NTCP are selected for proton therapy, the average NTCP equals $13.2 \%$. For the SEQ scheme, the average NTCP values for xerostomia equal $13.6 \%$ and $14.2 \%$ for the combined and the single-modality treatment, respectively. Figure 1c shows the corresponding proton slot allocation for the SEQ scheme, indicating that only 3 patients receive proton slots for the boost plan whereas most proton slots are used for base plans. To achieve an average NTCP of $14.2\%$, combined treatment would require only 265 ($16.8\% $) proton fractions instead of 315 ($20\% $). Similar results were obtained for NTCP models for dysphagia and aspiration.

Conclusion: Combined proton-photon treatments with optimized allocation of proton slots increase the benefit of proton therapy on the population level compared to single-modality treatments with optimal proton patient selection. However, the benefit is small for the SIB scheme. A larger benefit is observed for the sequential boost scheme, where combined treatments can exploit that some patients benefit from proton boost plans and others from proton base plans.

[1] A. Jakobi et al., Identification of Patient Benefit From Proton Therapy for Advanced Head and Neck Cancer Patients Based on Individual and Subgroup Normal Tissue Complication Probability Analysis, IJROBP, V. 92(5), pp.1165-74, 2015

Speaker: Mr Nicolas LOIZEAU (Department of Physics, University of Zürich, Switzerland)

13:30 Simultaneous optimization of photon, electron and proton beams

Introduction
Recent research has shown the feasibility to combine the respective dosimetric advantages of photon and electron beams to achieve superior treatment plan quality (mixed beam radiotherapy MBRT) in comparison to pure photon therapy. Tumor treatment with proton beams on the other hand has distinct benefits in terms of reduced integral dose to a patient compared to photon beams. The purpose of this work is to develop a new and potentially superior treatment modality, called triple beam radiation therapy (TBRT), to exploit the advantages of photon, electron and proton fields.

Materials and Methods
The Eclipse treatment planning system (version 15.6), is used to set up beam directions for photon and electron fields and proton spot candidates for TBRT plans. The Swiss Monte Carlo Plan (SMCP) is used to calculate beamlet dose distributions for all three particle types. In order to investigate the potential of TBRT plans, a 4PI field setup for photons and protons is chosen. For the electrons, a coplanar setup with source to surface distance between 70-90 cm is chosen. A fluence map optimization (FMO) is performed to simultaneously optimize all three particle types to generate a TBRT plan. These FMO optimized 4PI TBRT plans are compared to MBRT plans in terms of DVH parameters for two academic, a clinical head and neck and a pelvic case. The PTVs in the two academic cases are located partly superficial and reaching into depths of up to 10 cm. The clinical head and neck case contains a mostly superficial PTV. For the pelvic case, the PTV is located between 0.5 to 7.5 cm within the body.

Results
While allowing three instead of just two (electron and photon) beam types, superior plan quality is achieved throughout all studied clinically motivated and academic cases. TBRT plans achieve better sparing of organs at risk (OARs). For instance, in the clinical pelvic case, the mean dose received by the bladder and the bowel is 36% and 73% lower for TBRT than for MBRT, respectively. The PTV homogeneity HI95 (V95-V107) improves by 2% and HI98 (V98-V102) improves by 18% for the TBRT plan in comparison to the MBRT plan. In the TBRT plan, each beam type contributes substantially to the dose distribution in the target. The mean dose fraction of each beam type delivered to the PTV in studied clinical cases are 40.5% electrons, 11.1% photons and 48.4% protons for the pelvic case and 19.8% electrons, 50.6% photons and 29.6% protons for the head and neck case.

Conclusion
A framework to simultaneously optimize photon, electron and proton beams based on Monte Carlo calculated beamlet dose distributions was successfully developed. FMO optimized triple beam treatment plans show dosimetric advantages, especially in sparing OARs in comparison to MBRT plans. The research is partially supported by Varian Medical Systems.

Speaker: Hannes LÖBNER (Division of Medical Radiation Physics and Department of Radiation Oncology, Inselspital, Bern University Hospital and University of Bern, Switzerland)

13:45 The importance of the Monte Carlo physics settings for simulations in proton therapy

Introduction: The Christie NHS Foundation Trust began treating patients using proton therapy in Dec. 2018. As part of the current patient specific quality assurance, each proton field is delivered to a SolidWater (SW) phantom (1 hour for preparation/analysis per patient plus 2 hours of beam time per plan). Monte Carlo (MC) based independent dose calculations have been proposed to reduce these measurements, however, when implementing a multi-purpose Monte-Carlo code such as Geant4, the underlying physics settings have to be chosen from a wide range of possible options. We aim to characterize their influence on clinical dose calculations when taking the whole MC beam-modelling process into account.
Material and Methods: A GATE 8.1 (Geant4 10.3.3) based MC system was set-up for clinical dose calculations as follows: 1) Choose underlying Geant 4 settings, 2) define initial beam optics, 3) adjust energy and energy spread to reproduce proton integral depth dose curves, 4) model beam modifying devices (pre-absorber, 2/3/5 cm Lexan plates to treat superficial tumours). This process was repeated using two pre-built MC physics lists, which differ in the modelling of electro-magnetic interactions, namely QGSP_BIC (EM Opt 0) and QGSP_BIC_EMZ (EM Opt 4). Example simulation results showing spot sizes in air with and without a pre-absorber positioned 46 cm upstream of iso-center are presented in comparison to commissioning measurements. Finally, for one clinical plan (paediatric patient, sarcoma of the neck, 5cm pre-absorber) the influence of different physics settings is shown in the patient CT, and dose distributions simulated in a SW phantom are compared to patient specific quality assurance measurements (two sets of repeated measurements, analysed with 2%/2mm gamma analysis).
Results: Without pre-absorber, beam sizes (sigma) in air are marginally affected by the choice of physics settings (agreement between QGSP_BIC/QGSP_BIC_EMZ simulated spot sizes within 0.1mm, red lines in Figure 1a). Differences in simulated scattering are however relevant when a pre-absorber is inserted into the beam (green/blue/black lines in figure 1a, differences of up to 1 mm for a 5 cm Lexan pre-absorber), with QGSP_BIC_EMZ showing closer agreement to measurements than QGSP_BIC (0.24mm vs 0.72mm difference). The influence of this scattering difference is demonstrated for a clinical case in figure 1b and 1c. Agreement to patient specific quality assurance measurements is higher for QGSP_BIC_EMZ when compared to QGSP_BIC (100% vs. 98.2%at 1.3 cm depth, 99.5% vs. 99.0% at 4.3 cm depth (treatment room 1) and 94.8% vs 92.7% at 1.3 cm depth, 97.9% vs 97.7% at 4.3 cm depth (repeat measurements in treatment room 2). Calculation times are higher (factor of 1.6/1.4 in the CT/SW) for QGSP_BIC_EMZ.
Conclusion: First results indicate that in Geant4 10.3.3, QGSP_BIC_EMZ reproduces measurements more accurately when compared to QGSP_BIC for treatments with pre-absorber, which comes at the cost of increased calculation time. As such, MC simulations are a promising tool to reduce the amount of physical measurements for proton therapy, but it is crucial to carefully choose the underlying settings, as for example differences in electro-magnetic models included in pre-built Geant4 physics lists affect the scattering of clinical proton beams and lead to differences in simulated doses. This work forms the first part of a multi-institutional study which aims to establish recommendations for MC settings for proton therapy.

Acknowledgements: This work was funded by the Engineering and Physical Sciences Council [grant number EP/R023220/1] and the Science and Technology Facilities Council [grant number ST/N002423/1]. Supported by the NIHR Manchester Biomedical Research Council.

Speaker: Carla WINTERHALTER (1 Division of Cancer Sciences, University of Manchester, M13 9PL, Manchester, UK; 2 The Christie NHS Foundation Trust, M20 4BX, Manchester, UK)

14:00 Liver Ultrasound as a predictor for online proton 4D dose distributions in lung tumours

Introduction: Motion management is crucial when applying scanned proton therapy to lung tumours. In order to mitigate
the detrimental motion effects, it is important to know the deformable motion of the patient’s lungs during
treatment. To date, no real-time 3D imaging modality is available, which is why a surrogate for the motion
is needed. This study investigates the predictive power of liver ultrasound (US) images combined with a
statistical motion model for the proton dose distributions in the lung.

Materials & Methods: Liver ultrasound and lung 4DMR images of two volunteers were acquired simultaneously during 10 minutes of
free breathing, resulting in ~700 variable 3DMRI volumes per volunteer. The breathing motion was extracted
using deformable image registration. The deformation vector fields (DVFs) were used to warp full-exhale CTs
of two lung cancer patients, resulting in four sets of synthetic 4DCTs. After combining the geometry of both
patients with the motion of both volunteers, each synthetic dataset contained ~700 motion states, which were
considered as “ground truth”.
For each patient/volunteer combination, a statistical motion model based on Gaussian process regression was
trained on the first ~600 motion states, correlating the US images with the corresponding MRI DVFs. This
model was used to predict MRI DVFs from the remaining ~100 US images. These DVFs are referred to as the
“predicted motion”.
Two-field PBS proton treatment plans were optimised on the geometrical ITV, including all CTV positions of
one respiratory cycle. The planning CT was defined in the following way: the HU value per voxel outside of
the ITV was the average of all phases of this respiratory cycle, whereas within the ITV, maximum intensity
projection was applied.
For these plans, 4D dose distributions with varying starting phases were calculated for the ground truth as
well as for the predicted motion. The results were analysed in terms of absolute dose differences and dosedifference-
histograms in the CTV.

Results: Figure a) shows the dose-difference histograms for all four datasets, including 4DDCs depending on all starting
phases in the shaded areas, with median values shown by the solid line. The dose differences due to model
prediction errors are mostly within 10% of the prescribed dose (median Vdiff>10% = 0.4%, 5.5%, 2.8%, 2.6%),
with most of the voxels showing a difference of less than 5% (median Vdiff>5% = 11.5%, 30.6%, 14.3%, 13.4%).
An example comparison of dose distributions based on ground truth and predicted motion is shown in Figure
b).

Conclusion: Our study suggests that liver US in combination with a statistical motion model can accurately predict lung 4D
dose distributions. Such a framework is thus useful for providing online image guidance for real-time proton
beam adaptation. A further study will investigate the effectiveness of proton tracking for the lung based on
this model.

Speaker: Ms Miriam KRIEGER (PSI)

14:15 Impact of interfractional deformable image registration variations for proton therapy of lung cancer treatments

Introduction: With proton therapy, high dose conformity to the target can be achieved while sparing normal tissues, which makes it especially suitable for non-small cell lung cancer (NSCLC) patients. On the other hand, the proton dose is sensitive to density changes in the beam path and the anatomy of NSCLC patients often changes between fractions (fast tumor growth/shrinkage, weight changes). To evaluate the effect of these changes on the dose distribution, it is necessary to deform the recalculated dose distributions on regularly acquired repeated images via deformable image registration (DIR) to match the planning CT. However, uncertainties in DIR lead to differences in the dose distribution warped back to the planning CT, which can influence any further clinical decision. In this study, we aim to evaluate the dosimetric uncertainty introduced by DIR, and to compare its magnitude to the differences caused by anatomical changes.
Materials and Methods: For 7 NSCLC patients we designed proton treatment plans with 60 Gy-RBE to the PTV. We recalculated the dose on 9 repeated breath-hold CTs at different time points during treatment and warped these doses back to the planning breath-hold CT for accumulation with 6 clinically used DIR algorithms (see Figure1). Differences in PTV V95 were evaluated.
Results: Example accumulated dose distributions warped with different DIRs are shown in Figure1. Large dose differences caused by anatomical changes were also seen in the deformed doses. By choosing different DIRs, differences up to 20% in PTV V95 can be observed.
Conclusions: Dose differences caused by anatomical changes are generally larger than those caused by the use of DIR algorithm. However, DIR variations should be considered when taking clinical decisions.

Figure 1:Initial plan dose, accumulated doses warped with different DIRs and DVH of PTV, CTV and ipsilateral lung. The solid line is the initial planned dose. The band shows the variations between accumulated doses with different DIRs

Speaker: Ms Lena NENOFF (Paul Scherrer Institute, Center for Proton Therapy, Switzerland; Department of Physics, ETH Zurich, Switzerland)

14:30 → 14:45 Closing Remarks

14:45 → 15:00 Coffee break & Group organisation for the guided tour

15:00 → 16:00 Center for Proton Therapy - Guided tour