Alfredo Mirandola a) * Centro Nazionale di Adroterapia Oncologica (CNAO Foundation), Pavia 27100, Italy

Transcription

1 Characterization of a multilayer ionization chamber prototype for fast verification of relative depth ionization curves and spread-out-bragg-peaks in light ion beam therapy Alfredo Mirandola a) * Centro Nazionale di Adroterapia Oncologica (CNAO Foundation), Pavia 27100, Italy Giuseppe Magro* Centro Nazionale di Adroterapia Oncologica (CNAO Foundation), Pavia 27100, Italy Universita degli Studi di Milano, Milano 20100, Italy Marco Lavagno De.Tec.Tor. S.r.l., Torino 10144, Italy Andrea Mairani Centro Nazionale di Adroterapia Oncologica (CNAO Foundation), Pavia 27100, Italy Heidelberg Ion Beam Therapy Center (HIT), Heidelberg 69121, Germany Silvia Molinelli, Stefania Russo, Edoardo Mastella, and Alessandro Vai Centro Nazionale di Adroterapia Oncologica (CNAO Foundation), Pavia 27100, Italy Davide Maestri Centro Nazionale di Adroterapia Oncologica (CNAO Foundation), Pavia 27100, Italy Universita degli Studi di Milano, Milano 20100, Italy Vanessa La Rosa De.Tec.Tor. S.r.l., Torino 10144, Italy Mario Ciocca Centro Nazionale di Adroterapia Oncologica (CNAO Foundation), Pavia 27100, Italy (Received 17 October 2017; revised 22 February 2018; accepted for publication 3 March 2018; published 6 April 2018) Purpose: To dosimetrically characterize a multilayer ionization chamber (MLIC) prototype for quality assurance (QA) of pristine integral ionization curves (ICs) and spread-out-bragg-peaks (SOBPs) for scanning light ion beams. Methods: QUBE (De.Tec.Tor., Torino, Italy) is a modular detector designed for QA in particle therapy (PT). Its main module is a MLIC detector, able to evaluate particle beam relative depth ionization distributions at different beam energies and modulations. The charge collecting electrodes are made of aluminum, for a nominal water equivalent thickness (WET) of ~75 mm. The detector prototype was calibrated by acquiring the signals in the initial plateau region of a pristine BP and in terms of WET. Successively, it was characterized in terms of repeatability response, linearity, short-term stability and dose rate dependence. Beam-induced measurements of activation in terms of ambient dose equivalent rate were also performed. To increase the detector coarse native spatial resolution (~2.3 mm), several consecutive acquisitions with a set of certified mm-thick PMMA sheets (Goodfellow, Cambridge Limited, UK), placed in front of the QUBE mylar entrance window, were performed. The ICs/SOBPs were achieved as the result of the sum of the set of measurements, made up of a one-by-one PMMA layer acquisition. The newly obtained detector spatial resolution allowed the experimental measurements to be properly comparable against the reference curves acquired in water with the PTW Peakfinder. Furthermore, QUBE detector was modeled in the FLUKA Monte Carlo (MC) code following the technical design details and ICs/SOBPs were calculated. Results: Measurements showed a high repeatability: mean relative standard deviation within 0.5% for all channels and both particle types. Moreover, the detector response was linear with dose (R 2 > 0.998) and independent on the dose rate. The mean deviation over the channel-by-channel readout respect to the reference beam flux (100%) was equal to 0.7% (1.9%) for the 50% (20%) beam flux level. The short-term stability of the gain calibration was very satisfying for both particle types: the channel mean relative standard deviation was within 1% for all the acquisitions performed at different times. The ICs obtained with the MLIC QUBE at improved resolution satisfactorily matched both the MC simulations and the reference curves acquired with Peakfinder. Deviations from the reference values in terms of BP position, peak width and distal fall-off were submillimetric for both particle types in the whole investigated energy range. For modulated SOBPs, a 2266 Med. Phys. 45 (5), May /2018/45(5)/2266/ American Association of Physicists in Medicine 2266

2 2267 Mirandola et al.: MLIC prototype for QA in particle therapy 2267 submillimetric deviation was found when comparing both experimental MLIC QUBE data against the reference values and MC calculations. The relative dose deviations for the experimental MLIC QUBE acquisitions, with respect to Peakfinder data, ranged from ~1% to ~3.5%. Maximum value of 14.1 lsv/h was measured in contact with QUBE entrance window soon after a long irradiation with carbon ions. Conclusion: MLIC QUBE appears to be a promising detector for accurately measuring pristine ICs and SOBPs. A simple procedure to improve the intrinsic spatial resolution of the detector is proposed. Being the detector very accurate, precise, fast responding, and easy to handle, it is therefore well suited for daily checks in PT American Association of Physicists in Medicine [ doi.org/ /mp.12866] Key words: ion beam therapy, MLIC, QA 1. INTRODUCTION Charged particle beams are particularly effective in cancer treatment because of their physical and radiobiological properties. The depth dose distribution of a monoenergetic heavy charged particle beam, entering the patient body, is characterized by a relatively low dose in the entrance region with low linear energy transfer (LET), a sharply prominent dose in the proximity of the penetration range (with higher LET) and a steeply distal fall-off beyond the peak. This so-called inverse depth dose profile, with respect to photon radiation, allows absorbed dose to be minimized within proximal and distal healthy tissues. Moreover, the relative biological effectiveness (RBE) values of higher Z ion beams (carbon ions, in particular) enhance the therapeutic gain, when compared with photons and lower Z particle beams, such as protons. RBE varies with LET; therefore, it depends on the penetration depth along the ionization curve (IC). 1 Since particles deposit the dose in a relatively small volume and their penetration range is highly sensitive to even small variations of material density and stopping power along the beam path, minimizing every possible source of uncertainty is a crucial aspect. Differently from conventional photon radiotherapy, for ion beams, a wrong estimation of in-patient beam penetration might cause dramatic effects in terms of tumor underdosage and/or normal tissues overdosage. 2,3 As a first step, prior to the complex task of in vivo range verification, 4,5 it is crucial to measure the Bragg peak (BP) curves with high precision, accuracy, and reliability, 6 both during the beam commissioning phase and periodic quality assurance (QA). However, a careful check of the beam energy stability, as part of the daily QA program, could be too timeconsuming if carried out with motorized water phantoms. High spatial resolution and sensitivity detectors can be used for IC measurements (such as diodes or diamonds), 7,8 but the acquisition time would be very long for daily QA, particularly if more than one particle type needs to be checked. Radiochromic films in a solid slab phantom can be used for BP position measurements for proton pencil beams, with a very satisfactory spatial resolution. 9,10 This procedure was only validated for proton beams, while procedures based on double-wedge and scanning beams were proposed for carbon ion energy checks. 10,11 However, besides the strong energy dependence of radiochromic film response, 12,13 the use of films is again time-consuming, since they need to be scanned and analyzed offline with a dedicated scanner and software. Multilayer ionization chambers (MLIC) have been therefore proposed for fast range measurements for QA purposes, especially in case of proton beams In this work, we focused on evaluating the MLIC module of the QUBE 32 channels (De.Tec.Tor., Torino, Italy) detector, for range measurements in both scanning proton and carbon ion beams. The full experimental dosimetric characterization of the detector, together with its detailed implementation into a Monte Carlo (MC) code, has been here reported, for the first time, for the two different particle types. 2. MATERIALS AND METHODS All measurements were performed at the Italian National Center for Oncological Hadronterapy (CNAO) synchrotron facility, in one of the treatment rooms equipped with a horizontal fixed beam line and pencil beam scanning (PBS) as treatment delivery modality. 20 Some representative beam energies in the whole clinical available range ( MeV and MeV/u for protons and carbon ions, respectively) were investigated for the detector characterization. The spot dimension, in terms of full width half maximum at the isocenter in air, varies as a function of the beam energy, from 2.2 to 0.7 cm and from 0.8 to 0.4 cm for protons and carbon ions respectively. 9 2.A. The MLIC QUBE detector QUBE is an in air-vented device developed for QA checks in particle beams, including three different modules in the full commercial version. The main module is composed of a MLIC detector, designed to measure the relative depth ionization distribution at different beam energies and modulations in depth. The strip chambers and the pixel chamber, used for beam position and spot dimension checks, are two optional modules that can be added on request. In this study only the MLIC main module was investigated.

3 2268 Mirandola et al.: MLIC prototype for QA in particle therapy 2268 The standard version of the MLIC QUBE module is composed of 128 channels; for this study a prototype, composed of 32 parallel plate ionization chambers, stacked in series, was used. In each chamber, 1-mm-thick aluminum plate is used as beam energy absorber and cathode. Each chamber has a sensitive area of mm 2. MLIC detector is intended for measurements of SOPB and pristine peak penetration ranges; it is not suitable for patientspecific field measurements for two main reasons: (a) the lateral dimensions of the sensitive area may not be large enough for collecting the whole ionization produced by the patientspecific field and (b) each integral plate collects the charge deposited on its whole area by all the traversing pencil beams; therefore, the device is not able to distinguish both the position and the penetration range of 2D spots for the delivered and potentially inhomogeneous patient-specific fields. The QUBE detector datasheet states that both the prototype and the standard version properly work when the operating environment ensures the humidity level between 5% and 60%, the temperature between 15 C and 30 C and the pressure range between 700 and 1060 hpa. The small size ( mm 3 ) and weight (8 kg) of the prototype make the device easy to be placed on the treatment couch. The beam passes through its mylar thin entrance window and a set of homogenous absorbing electrodes, each separated by an air gap. The ionization charge is collected along the beam penetration axis. Relative depth ionization distributions can be therefore instantaneously measured, due to the simultaneous readout of all the individual chambers. Further cross-calibration with water phantom measurements is necessary for absolute range evaluation. Both pristine and spread-out-bragg-peaks (SOBP) can be determined with a native nominal water equivalent resolution, provided by the constructor, of ~2.3 mm and therefore giving to this MLIC module a total water equivalent thickness (WET) of ~75 mm. The charge collected in the ionization chambers is converted into counts by the front-end board 21 and then sent via Transistor-Transistor-Logic protocol to a controller, for data communication. Differently from other types of MLICs, built using water equivalent channels (graphite electrodes on PMMA absorbers), 14,19 the electrodes of the MLIC investigated in this work are made of aluminum, which causes a particle scattering effect more similar to water compared to typical solid water materials A.1. Calibrations and short-term gain stability For a full characterization of the detector s materials, the thickness of a sample cathode was measured with a caliper at different positions (at the center and at the four corners of the slab). In order to perform a physical calibration of the detector, measurements of WET using the Peakfinder (PTW, Freiburg, Germany), were also carried out with both proton and carbon ion monoenergetic beams. Results achieved for the analyzed sample have been generalized to all cathodes. The ionization signal intensity for each channel is linear with the air gap through appropriately designed spacers. Hence, the admitted construction tolerance and uniformity of the spacers determine how the ionization signal is going to vary with respect to the expected value, and therefore the gain of each channel. To correct for such a variation, a gain calibration was performed before each measurement session, by acquiring high-energy ICs ( MeV for proton and MeV/u for carbon ions, whose BP depth in water corresponds to 250 and 270 mm, respectively). This ensured the MLIC module to acquire the entrance plateau region of the IC, which is approximately flat in water slightly increasing in depth and showing an initial buildup region, in particular for high-energy proton beams. 23,24 Referring to the nominal IC in water, in fact, the ratio between the value at around 75 mm in depth (i.e., the MLIC QUBE s WET) and the entrance point, is about 1.09 and 1.03 for MeV proton beam and MeV/u carbon ion beam, respectively. These channel-specific calibration factors were applied for all the measurements and MC simulations described in the following. Values of the characteristic depths and lengths (see Section (2.C)) of the ICs are not affected by the presence of minor inaccuracies in the first channels readout, which may be found by neglecting the buildup region of the applied gain calibration curve. The short-term stability of the gain calibration was assessed, for both particle types, by re-delivering the reference beam energies 1 and 3 h after the reference gain calibration acquisition. Short-term stability was assessed even in case of stress test (power supply unplugged/plugged, cables disconnected/connected, high voltage off/on). Regarding the long-term stability, the manufacturer recommends to perform a gain re-calibration at the beginning of each measurements session. Same procedure is suggested for similar commercial equipment like Zebra and Giraffe (IBA Dosimetry GmbH, Schwarzenbruck, Germany). 2.A.2. SOBP plans To assess the dependence of the MLIC QUBE both on particle type and beam energy, the detector was also tested with a set of cubic volumes of known dose, optimized in water, covering the whole available energy range. Three different cubes of cm 3, centered at 6.5, 13.5, and 24.5 cm in water, were generated using our clinical treatment planning system (TPS), Syngo RT V13 (Siemens AG, Healthcare Sector, Erlangen, Germany), for both particle species. However, a flat SOBP profile measured with the MLIC QUBE does not necessarily represent a flat depth dose distribution since the charge collected on the integral planes is not proportional to dose, but rather to dose-area product. Nevertheless, the chosen lateral dimensions (3 cm) represent a reasonable value for having a homogeneous dose distribution (both in lateral and longitudinal directions) for the volumes optimized by our TPS. Moreover, the limited dimensions of the cubic volumes, when compared with the plate area ( mm 2 ), realistically allow the deposited

4 2269 Mirandola et al.: MLIC prototype for QA in particle therapy 2269 ionization to be collected totally. For the shallowest target, a 3-cm-thick solid water (RW3, PTW Freiburg, Germany) range shifter was added. The dose distribution was optimized to be uniform in terms of physical dose for proton radiation and in terms of (local effect model-based as presented in Ref. [25]) biological dose for carbon ion radiation. High dose homogeneity was required both in the lateral and in the longitudinal directions, by adopting the clinical configuration used for patients treatments. In particular, 3 mm spot spacing was used for protons, while 2 mm spot spacing and two ripple filters 9,26 were used for carbon ions. The iso-energy slice (IES) spacing in depth was set to 2 mm for both particle types, so that each cubic volume was made of 16 IESs. 2.B. Experimental setup and measurements The experimental apparatus is shown in Fig. 1. In order to measure both pristine BPs and SOBPs in a single-shot acquisition, 32-channel QUBE prototype was aligned with the entrance window of the MLIC module at the isocenter. The mylar entrance window of the QUBE was aligned at the room isocenter and clinical beam line settings were adopted for the irradiation: two 2-mm-thick ripple filters for carbon ions, while no energy modifiers were used for protons. To explore the whole clinical energy range available by the synchrotron, a representative sample of energies has been delivered: 81.56, , , and MeV for protons while , , , , and MeV/u for carbon ions. The previous listed beam energy corresponds to BP depths in water of 50, 101, 151, and 250 mm for protons and 50, 100, 150, 200, and 250 mm for carbon ions, respectively. About and particles were delivered within five consecutive spills, for protons and carbon ions, respectively. In both cases, the delivery time needed for each measurement was less than 20 s, including the interspill pauses in which the beam is not available. Because of the reduced longitudinal size of the QUBE prototype used for this study (32 channels), a variable thickness of absorbing material (RW3) was needed to properly shift a subset of Bragg curves and SOBPs within the detector water equivalent measuring range (~75 mm). Table I summarizes the experimental setup for each measurement session. FIG. 1. Experimental setup: 32-channel QUBE. [Color figure can be viewed at wileyonlinelibrary.com] Same settings were adopted with the PTW Peakfinder water column detector, 9,11 here used as a reference, because of its superior performance in terms of spatial resolution (0.05 mm around the BP for our work). Once high acquisition repeatability was experimentally assessed (see Section (3.B)), each IC (both BPs and SOBPs) was then obtained by merging the subset of measurements made up of a one-by-one PMMA layer acquisition (each certified PMMA sheet (Goodfellow, Cambridge Limited, UK) was mm thick). 2.B.1. Short-term repeatability The acquisitions repeatability represents an essential prerequisite for the usage of the QUBE in particle beams QA. As described in the above, the procedure of increasing the intrinsic spatial resolution of the detector also requires a test retest reliability. Therefore, to assess the repeatability of the measurements under proton and carbon ion beams, nine consecutive acquisitions of monoenergetic pencil beams were performed. The beam energies investigated for this part of the study were MeV for protons and MeV/u for carbon ions. About and particles were delivered within four spills, for protons and carbon ions, respectively. In both cases, the delivery time needed for each measurement was around 10 s, including the interspill pauses in which the beam is not available. 2.B.2. Linearity with dose and dose rate dependence Given the fact that the QUBE detector is designed for fast daily QA measurements, it is of paramount importance for TABLE I. List of proton (p) and carbon ion (C) beam types and experimental setup used for the characterization of the MLIC QUBE detector prototype. Beam type Monoenergetic pencil beam (p) Monoenergetic pencil beam (C) mm ripple filters Modulated scanning beam, SOBP (p) Modulated scanning beam, SOBP (C) mm ripple filters Nominal BP position/nominal center of the MR in water [mm] Range shifter (RW3) Thickness [mm] SOBP, spread-out-bragg-peak; BP, Bragg peak; MR, modulation region, the nominal extent of the flat physical dose region.

5 2270 Mirandola et al.: MLIC prototype for QA in particle therapy 2270 the device to perform linearly with dose and with same accuracy for all beam flux levels (i.e., number of particles per second), provided by the synchrotron and commissioned for clinical use, so not being prone to day by day possible intensity variations. Linearity with dose was investigated for the reference dose rate and beam energy ( MeV/u for carbon ions and MeV for protons) at several clinically relevant beam intensities, ranging from to particles for carbon ions and from to particles for protons. These intensity values were suitable for a fast QA (from about 1 s as to 10 s of delivery time). In order to assess the dependence on the dose rate, a MeV/u monoenergetic carbon ion pencil beam of a fixed beam intensity (total number of particles of ) was delivered with three machine different configurations, as available from our synchrotron: 100%, 50%, and 20% of the full beam. The 100% of beam flux has been taken as a reference for the other beam levels, which are achieved with the insertion of intensity degraders acting as beam passing filters in the extraction line B.3. Range resolution Given the detector specifications, the available number of channels and the water equivalent measuring range, the intrinsic native spatial resolution of the MLIC QUBE is quite coarse (around 2.3 mm). Hence, it is neither enough to precisely determine the position of the BP in depth nor to fulfill the accuracy requested by our QA protocol, 9 particularly for low-energy beams with the narrower BP width. A direct comparison between the IC acquired with the MLIC QUBE detector and those acquired with the Peakfinder cannot be easily performed, both due to a different spatial resolution and medium which the particle beams interact with (mainly aluminum for the MLIC QUBE and water for the Peakfinder). Therefore, some procedures or methods, at least to increase the spatial resolution of the detector under study, should be implemented. Nevertheless, differences of the detectors material may affect the peak-to-entrance ratio of the acquired ionization profiles, being more pronounced for higher energies and heavier ions, for which nuclear interaction processes dominate. The former task was successfully accomplished by performing several acquisitions in the same conditions, both in terms of number of particles delivered and dose rate, with a series of thin PMMA layers (WET of 0.22 mm each) placed in front of the QUBE mylar entrance window. This way potentially misleading identification of the maximum signal channel with the one of reference for the BP position, due to the obvious limitation of the sampling procedure, is prevented. 2.B.4. Radiation-induced device activation In order to give an indication about the potential risks to the operators handling the QUBE detector, an evaluation of the device activation was performed after the irradiations. Measurements were carried out 50 cm away from the detector (roughly representing the distance which the operators work at) and at the entrance window (representing the highest activation area). The induced radioactivity was assessed with a plastic scintillator model AT1123 (ATOMTEX, Belarus) expressed in terms of ambient dose equivalent rate, _H ð10þ, with an uncertainty of around 5%. The measurements were performed for both short (about 110 s with and particles for carbon ions and protons, respectively) and long acquisition of about 20 min with particles delivered with a MeV/u carbon ion beam. The latter was chosen to assess radioactive contamination in a high dose scenario. In addition, for the long acquisition case, a comparative evaluation on _H ð10þ was performed in the same experimental condition with the PTW Peakfinder and the IBA Giraffe detector. 2.C. Data analysis Aiming at a full characterization the MLIC QUBE detector, the BP position here defined as the mean value between the 95% proximal and distal of the maximum dose level has been evaluated. Then, a deeper investigation of the experimental pristine ICs was carried out by evaluating the following quantities for both the MLIC QUBE and the MC simulations: BP full width at 50% (70%) of the maximum (FW50% M, FW70%M), defined as the distance between the 50% (70%) distal and proximal dose levels; Dose distal fall-off (DDF), 28 defined as the distance between the 80% distal and the 20% distal of the IC, normalized to the maximum dose level; The practical range (R p ), 28 defined as the depth, distal to the BP, at which the dose is 10% of its maximum value; Additionally, for carbon ion beams, a further analysis was performed about the fragmentation tail. The integral dose between the 110% of the position of maximum ionization value and the last available data point in depth has been computed and presented in terms of percentage difference with respect to the Peakfinder calculated value. For the analysis of the SOBPs the following parameters have been assessed: Protons: Range equivalent depth 28 difference (DR): the difference between the MLIC QUBE (both experimental and MC) and the Peakfinder measured position of the distal 80% dose of the SOBP normalized at the center of the modulation region (MR, the nominal extent of the flat physical dose region); R p as defined in the above, with the exception of the normalization point, now taken at the center of the MR;

6 2271 Mirandola et al.: MLIC prototype for QA in particle therapy 2271 SOBP width defined as the distance between the distal 90% and the proximal 90% dose normalized at the center of the MR; DDF as the distance between the distal 20% and 80% dose normalized at the center of the MR. Carbon ions: Depth of the distal 50% of the dose normalized at the center of the MR (the nominal extent of the corresponding flat biologically optimized dose region); as an estimate of the particle range; DDF as the distance between the distal 30% and 70% dose normalized at the center of the MR Depth dose profiles, normalized at the center of MR, were further analyzed by taking the average of the absolute values of the relative dose differences, between Peakfinder (reference) and MLIC QUBE/MC data, along the curve. This served as an evaluator of the dosimetric agreement. In the case of carbon ions, a slightly different analysis has been performed. The physical dose related to a LEM-optimized SOBP is obviously not flat as a function of depth in water, and the same analysis proposed for proton SOBPs cannot be easily adapted to this context. Moreover, the relatively high-dose level of the measured fragmentation tails prevented the low-dose points (i.e., distal 10% or 20%) to be considered as good indicators for both the DDF and the particle range, as defined in Ref. [28]. Unless otherwise stated, results have been presented as differences with respect to the Peakfinder measured values, taken as reference. The measured ICs, here investigated for the repeatability, short-term stability of the gain calibration, linearity with dose and dose rate dependence assessment, were all normalized to the reading of the first channel of the detector. A threshold of 3% of the maximum measured value of each acquisition was additionally applied to exclude those channels readings which could be potentially affected by readout fluctuations at very low-dose levels. Considering repeatability and short-term stability analysis, results have been presented in terms of channel mean relative standard deviation (i.e., mean coefficient of variation), defined as the average value of the ratios between the standard deviation and mean readouts for each channel of the MLIC QUBE. The evaluation was performed over nine consecutive measurements for the repeatability assessment and over three acquisitions at different times for the short-term stability assessment. In order to quantify the detector linearity with dose, the channels readouts were analyzed as a function of five beam intensity levels (number of particles) for both protons and carbon ion beams of a fixed energy, as reported in Section (2.B.2). The R 2 index was then used as a goodnessof-fit statistics for the linear fit. For the dose rate dependence, an average value over the channel-by-channel deviation to the reference beam flux (100%) was calculated for both the 50% and the 20% of the full beam. For the sake of completeness, minimum and maximum values of the averaged interval have been reported for the repeatability, short-term gain stability and dose rate dependence measurements. For the SOBPs analysis, a dose threshold criterion (1.5% of the highest dose value) coupled with a gradient-based (0.5 a.u./mm) exclusion criterion have been applied to exclude values potentially affected by readout fluctuations at lowdose levels and points lying on steep dose gradients (i.e., points on the DDF), which may cause major deviations between the two datasets under comparison. 2.C.1. MC validation By following the detector design details, both mylar thin entrance window and each electrode were modeled with all its constituents to cover its nominal thickness of 12 lm and 1.16 mm, respectively. Short preliminary run allowed to slightly adjusting the MC input parameter for the aluminum ionization potential (I) to obtain the average value for the WET as measured experimentally. As representative energies, to retrieve a suitable value for I, a MeV (BP at 151 mm) and a MeV/u (BP at 200 mm) proton and carbon ion pencil beams were used. The null shift of the BP position reported in Table III for these two energies is a proof of the described procedure. As already specified in Section (2.C), all the parameters listed for the analysis of BPs and SOBPs have been also extracted from the MC outcomes. For a complete cross-validation of the MC simulation of the MLIC QUBE detector, a further analysis has been then performed by evaluating the mean of the absolute values of the local dose differences between MC and MLIC QUBE data, normalized to the experimental value. To avoid misleading results, again a dose threshold (1.5% of the maximum dose value) and a dose-gradient (0.5 a.u./mm) exclusion criteria have been applied. The percentage of remaining data (1-g) was used as normalization factor for this mean absolute deviation (Adj. MAD, Eq.(1)) in order to make an un-biased overall evaluation of the agreement between MC and experimental data: Adj:MAD ½%Š ¼ 1 1 g 1 X di MC di DATA (1) N i d DATA i where g is the percentage of data satisfying the exclusion criteria reported in the above, d i.is the dose at i-th point for MC and experimental data and N is the total number of point doses. 3. RESULTS 3.A. Calibrations and short-term gain stability The resulting thickness of the sample cathode was mm, thus confirming that the manufacturing process, the coupling of the material components and the

7 2272 Mirandola et al.: MLIC prototype for QA in particle therapy 2272 gluing has been done very accurately. As far as WET measuring is concerned, slight differences have been found between proton and carbon ions when using the same set of energies as listed in Section (2.C). In particular, the relative water equivalent path length (rwepl, 9 ) of the sample cathode ranged from 1.96 to 2.02 for protons, with the increasing energy. Higher values of rwepl were measured for carbon ions, ranging from 2.10 to The WET of the thin mylar entrance window was 30 lm for both particle types. As a result of the gain calibration process (see subsection (2.A.1)), the detector readout was verified to be in full agreement with reference calibration IC in water, channel-by-channel, within 1%. The short-term stability of the gain calibration was assessed for both particle types by comparing, channel by channel, three gain calibration acquisitions, delivered at different times (see subsection (2.A.1)). Results show a very good stability: the mean relative standard deviation of the channels readout was less than 1% for both particle types. As an indication, in case of the proton gain calibration, the relative standard deviation ranged from 0.2% to 0.7%, with a mean of 0.4%. 3.B. Repeatability, linearity with dose, and dose rate dependence As reported in Table II, the agreement among the nine runs performed is very satisfactory for both particle species, being the mean relative standard deviation of the channels readout less than 0.5%. Moreover, this value is ranging from about 0% to 1%, which can be a further assessment that all the channels well perform independently of their position relative to the Bragg curve. The linearity with dose was assessed by delivering, at a fixed dose rate, the same beam energy at five different beam intensities (number of particles delivered) for both proton and carbon ion beams. As an example, in case of carbon ions, Fig. 2 shows how the response of each channel, normalized to the first channel reading, is independent on the investigated beam intensity range. The channel reading as a function of number of particles is reported in the upper left panel of the Fig. 2, for a representative channel (#15). A good linear behavior was TABLE II. Channel mean, minimum, and maximum relative deviations for a proton (p) and a carbon ion (C) monoenergetic beam, specified by their nominal Bragg peak position in water. Beam type Nominal BP position [mm] Channel mean relative standard deviation (min.; max.) [%] Monoenergetic pencil beam (p) (0.1; 0.9) Monoenergetic pencil beam (C) (0.0; 1.2) mm ripple filters Abbreviations as defined in Table I. observed, with the R 2 of the linear best fit >0.998 for all the 32 channels of the prototype, for any particle type. The MLIC QUBE reading has been shown to be independent also on the dose rate, since the mean dose deviation with respect to the reference 100% beam flux level was 0.7% (ranging from 0.2% to 2.6%) and 1.9% (ranging from 0.3% to 6.0%) for the 50% and the 20% of the full beam, respectively. A deeper analysis on the electronics of the detector showed how the largest deviations (i.e., 2.6% and 6.0%) were attributed, for both cases, to a single defective channel. By dropping out these data, the final mean deviations remain substantially unchanged. 3.C. Procedure for range resolution improvement As introduced in Section (2.B.3), the spatial resolution of the detector has been increased up to about 0.5 mm by adopting thin plastic layers of well-known WET. As an example, Fig. 3 shows a convolution of four ICs (corresponding to the insertion of none, 1, 3 and 5 PMMA layers) as measured by the MLIC QUBE, for which the range resolution has been dramatically increased with respect to a single acquisition. The resulting superposition well reproduces the main features (for instance, BP position and DDF) of the extremely resolved Peakfinder-based acquisition (see the dotted line). At this stage, each layer was manually inserted, which explains the choice of four experimental configurations (see Fig. 3 and followings sections) as a good compromise between the number of acquisitions (i.e., detector s runs) and an acceptable spatial resolution. Since an optimization of the detector is ongoing, the range resolution could still be possibly improved with a QUBE-specific add-on, which automatically inserts the thin PMMA layers in front of the entrance window. 3.C.1. Monoenergetic pencil beams Figure 4 shows the comparison among the Peakfinder reference curves, MLIC QUBE data and the MC simulation, reproducing the same experimental setup. As reported in Table III, all parameters characterizing the ICs show a very good agreement with the reference values for both particle types in the full range of the investigated energies. For protons at MeV a discrepancy, in terms of plateau to peak ratio, is observed between Peakfinder and QUBE/MC data. This is probably due to the presence of a thick range shifter (140 mm thick for this specific case) that degrades the beam energy and considerably increases the scatter effect. Except for the FW50%M of carbon ions at 100 mm in water, only submillimeter differences can be noticed for all the peak widths. On average, a mm shift and a mm shift was registered for experimental nominal BP positions and both DDF and practical range, respectively. Similar results were found for MC. Even the integral fragmentation tails (see Section (2.C)), acquired either with the MLIC

8 2273 Mirandola et al.: MLIC prototype for QA in particle therapy 2273 FIG. 2. Monoenergetic MeV/u carbon ion pencil beam passing through mm ripple filters. Five different Bragg curves are shown with different markers depending on the beam intensity as displayed in the legend. Relative ionization for each channel is displayed for the five acquisitions with the MLIC QUBE. Each dataset has been normalized to the first channel reading. The upper left panel shows the linear behavior of one representative channel (#15). [Color figure can be viewed at wileyonlinelibrary.com] FIG. 3. Monoenergetic MeV/u carbon ion pencil beam passing through mm ripple filters +140-mm-thick RW3 range shifter. The Bragg curve is shown with a dotted line, acquired by the Peakfinder. The dots indicate four different MLIC QUBE configurations: none, one, three, and five plastic layers in front of the mylar entrance window. Each dataset has been shifted backwards by the water equivalent thickness of the added PMMA layers. The upper right panel shows a magnification of the peak region. The curves are normalized at the lowest depth achievable with the Peakfinder (13.4 mm). [Color figure can be viewed at wileyonlinelibrary.com] QUBE or simulated with MC, satisfactorily match the reference data: deviations were less than 1%. For the highest carbon ion beam energy, a slight larger deviation was found in the measured fragmentation tail: this may be due to the presence of a thick range shifter which worsens the beam quality. simulations against the Peakfinder reference values. Dose absolute deviations for the experimental MLIC QUBE acquisitions were within 1% to 3.5%. Nevertheless, they are still in close agreement with the MC, thus leading to conclude that MC outcomes may be independent of the modeling accuracy. 3.C.2. Modulated scanning beams (SOBPs) Figure 5 shows the comparison among the Peakfinder reference data, MLIC QUBE acquisitions and MC simulations reproducing the same experimental setup. On the whole, the MLIC QUBE measured values match the overall SOBP shapes and agree with the reference data satisfactorily, as well as the MC predicted values. The results concerning the parameters characterizing the SOBPs are listed in Table IV. On average, a submillimetric deviation was found when comparing both experimental MLIC QUBE data and MC 3.D. MC simulations As a further validation of the MC code, a specific analysis was performed between MLIC QUBE data and MC predictions. The adjusted MAD (see Eq. (1)) has been computed as described in subsection (2.C.1) and results are reported in Table V. The MC predicted values for the IC, which were verified channel by channel, show discrepancy lower than 5.6%. This is an acceptable value considering how the adjusted MAD may overestimate the potential disagreement between the

9 2274 Mirandola et al.: MLIC prototype for QA in particle therapy 2274 FIG. 4. Monoenergetic ionization curves as measured by the Peakfinder (dashed lines) compared against MLIC QUBE data (squares) and MC simulation (circles). Proton and carbon ion data are shown in the upper and lower panels, respectively. [Color figure can be viewed at wileyonlinelibrary.com] TABLE III. Outcomes of the analysis performed on the monoenergetic proton (p) and carbon ion (C) beams, specified by their nominal Bragg peak position in water. Results for MLIC QUBE data (A) and MC simulation (B) about the Bragg peak (BP) position, the full width at 50% and 70% of the maximum (FW50% M, FW70%M), the dose distal fall-off (DDF), the practical range a (R p, 28 ) are shown as differences (D) with respect to the Peakfinder measured values, taken as reference. Mean (l) and standard deviation (r) is reported for each column. In addition, for carbon ion beams, a further analysis has been performed on the fragmentation tail, as explained in Section (2.C). Particle type Nominal BP position [mm] DBP [mm] DFW50%M [mm] DFW70%M [mm] DDDF [mm] DR p [mm] Frag. tail dose relative dev.[%] A B A B A B A B A B A B p p p p C C C C a 0.3 a C a 0.2 a l r Abbreviations as defined in Table I and Section (2.C). a The relatively high-dose level of the measured fragmentation tails prevented the 10% of the maximum dose level to be adopted as indicators for R p. 28 The depth corresponding to the 20% of the maximum dose level was used instead. computed and the experimental data. The limiting factors to a better overall agreement (see Figs. 4 and 5) are given by the absolute value of the evaluated differences and the normalization factor taking into account the number points not matching the exclusion criteria described in subsection (2.C.1). As a single and representative unbiased parameter, the average value for the adjusted MAD has been then computed and found to be %, which is not far from the average statistical uncertainty affecting MC results (from 1.2% to 2.5% at the channel level). The agreement is therefore satisfactory. 3.E. Device activation For the beam intensities delivered during each of the short acquisition, _H ð10þ was found to be <130 nsv/h 50 cm away from the detector and <2.5 lsv/h at the entrance window. At the end of the long acquisition, the detected _H ð10þ was

10 2275 Mirandola et al.: MLIC prototype for QA in particle therapy 2275 FIG. 5. SOBPs as measured by the Peakfinder (black stars with dashed lines for visual guide only), compared against MLIC QUBE data (squares) and MC simulation (circles). Proton and carbon ion data are shown in the upper and lower panels, respectively, and normalized to the entrance point. The centers of the modulation regions are reported in terms of their nominal depth in water. [Color figure can be viewed at wileyonlinelibrary.com] TABLE IV. Outcomes of the analysis performed on the modulated scanning proton (p) and carbon ion (C) beams, specified by the nominal center of the modulation region in water. Results for MLIC QUBE data (A) and MC simulation (B) about the SOBP width, the range equivalent depth a (R, 28 ), the practical range (R p, 28 ), and the dose distal fall-off (DDF) are shown as differences (D) with respect to the Peakfinder measured values, taken as reference. Mean (l) and standard deviation (r) is reported for each column. In addition, a further analysis has been performed point-by-point to quantitatively assess the global agreement, on the whole acquired dose profile, as explained in Section (2.C). Particle type Nominal center of MR in water [mm] Dwidth [mm] DR [mm] DR p [mm] DDDF [mm] Absolute mean local charge percentage dev. [%] A B A B A B A B A B p p p C a 0.2 a C a 0.2 a C a 0.1 a l r Abbreviations as defined in Table I and Section (2.C). a The relatively high-dose level of the measured fragmentation tails prevented the 10% of the dose at the center of the MR to be assessed as indicators for R p. 28 The depth corresponding to the 50% of the dose at the center of the MR was used instead lsv/h at the entrance window and 0.3 lsv/h 50 cm away from the detector. As a method of comparison, the same long acquisition was performed with the Peakfinder and the Giraffe detectors. _H ð10þ values were 10.8 and 27.3 lsv/h at the entrance window, while an _H ð10þ of 0.5 and 0.4 lsv/h was detected 50 cm away for Peakfinder and Giraffe, respectively. For the long acquisition with the QUBE, some measurements were also performed at different times: 25 min after irradiation, _H ð10þ decreased to 3.5 lsv/h (entrance) and 150 nsv/h (50 cm away); after 20 h, _H ð10þ dropped everywhere to the natural background radiation level (~8090 nsv/h in the treatment room). 4. DISCUSSION For a full characterization of the detector, both the physical and gain calibration are mandatory prerequisites: details on these procedures are reported in subsection (2.A.1). For a

11 2276 Mirandola et al.: MLIC prototype for QA in particle therapy 2276 TABLE V. Results of the analysis performed on the pristine and modulated scanning proton (p) and carbon ion (C) beams, specified by the nominal Bragg peak position or the center of the modulation region in water. As explained in Section (2.C.1), the adjusted MAD has been computed. Mean (l) and standard deviation (r) is reported for each column. Beam type Monoenergetic pencil beam (p) Monoenergetic pencil beam (C) mm ripple filters Modulated scanning beam, SOBP (p) Modulated scanning beam, SOBP (C) mm ripple filters Nominal BP position/nominal center of MR in water [mm] Adj. MAD [%] (Eq. (1)) l 3.5 r 1.2 Adj. MAD, adjusted mean absolute deviation (see Eq. (1)), other abbreviations as defined in Table I. longer version of the QUBE and beam energies not high enough to traverse at the same time all the available channels with the entrance plateau region, a two-steps gain calibration procedure could be adopted. The detector, in fact, can be flipped 180, so even those channels not hit by the beam plateau during the first acquisition will be calibrated at the second run. Moreover, the detector is a modular device, so the number of channels can be adapted to the customer s needs. The first part of this study was dedicated to test the shortterm repeatability, linearity with dose and the dose rate dependence of the detector. To this end, several ICs have been acquired with the MLIC QUBE detector prototype and a high repeatability, linearity and dose rate independence have been proved. Once repeatability was assessed, an important part of this work was addressed to increase the coarse spatial resolution of the native detector. Both low-energy proton and carbon ion beams show a very narrow pristine Bragg peak, even less than 1.5 mm width. 9 For these critical conditions, in particular for carbon ions, the commonly used fitting approximation 29 could not perform accurately enough. This entails that increasing the native spatial resolution is highly recommended. Alternative to the more sophisticated method recently presented by, 30 based on a MC-generated lookup function, the procedure adopted in this study is direct and practical, being based on the insertion of thin PMMA sheets, acting as range modifiers, in front of the mylar entrance window of the detector. In principle, the spatial resolution can be increased depending on the number and the thickness of adopted layers. At this stage, each layer was manually inserted, which explains the choice of four experimental configurations as a good compromise among number of acquisitions, acceptable spatial resolution and delivery time. The convolution of the measured curves showed, even for a limited number of PMMA sheets, a spatial resolution around 0.5 mm, enough for the energy checks. Due to the limited size of the prototype, the insertion of thick range shifters (up to 20 cm) in front of the detector was needed to collect the full ICs. Although passive absorbers slightly perturbs the shape of the IC, particularly at the distal end, all the parameters characterizing the IC (both pristine BP and SOBPs), including the amount of dose delivered in the carbon ion fragmentation tails, were found to be in agreement with the reference values and/or within the tolerance levels established in our QA protocol. Furthermore, the full detector geometry and the elemental composition of all its constituents have been implemented within the FLUKA MC code. 31,32 Results about the comparison of the MC predictions and experimental measurements were successfully used as a validation of the detector modeling. The latter will be used for further developments of the QUBE detector. The acquisition time of the MLIC QUBE detector was short compared to the Peakfinder s one: a couple of seconds are enough to obtain each pristine IC with a satisfying signal to noise ratio (~1000). Several minutes are instead required for highly resolved data acquisition with the Peakfinder. The measurements of QUBE activation showed that no specific precaution has to be adopted by the operators when handling the QUBE device after irradiation sessions. However, as a general rule for all devices irradiated with ion beams, the exposure time due to handling should be minimized and the device itself properly stored after usage. In particular, at the working distance, _H ð10þ is slightly higher than the natural background radiation level, even immediately after the irradiation. In the unusual case of a very long acquisition, few minutes are enough to make the _H ð10þ values dropping rapidly to very low activation levels. In the latter case the induced radioactivity is about 50% less for QUBE when compared with a similar detector like the IBA Giraffe. Conversely, slightly lower exposure levels were found with PTW Peakfinder (at least at the entrance window) respect to the QUBE. Nevertheless, the Peakfinder is designed for much longer acquisitions, compared to the QUBE, being the former a scanning single-channel detector type. In addition, QUBE is easy to handle and to place on the treatment couch and few connection cables are needed. The full characterization of the MLIC detector prototype QUBE De.Tec.Tor. for the acquisition of ICs and SOBPs was carried out, for the first time, for both proton and carbon ion scanning beams from a fixed horizontal beam line. The dosimetric characterization here presented can be extended to the full commercial version of the device, already available on the market. Furthermore, this detector could be used for different beam line geometries (oblique or vertical) and for passive delivery technique. 33 Nevertheless, in case of