The first probe of a FLASH proton beam by PET

The recently observed FLASH effect related to high doses delivered with high rates has the potential to revolutionize radiation cancer therapy if promising results are confirmed and an underlying mechanism understood. Comprehensive measurements are essential to elucidate the phenomenon. We report the first-ever demonstration of measurements of successive in-spill and post-spill emissions of gammas arising from irradiations by a FLASH proton beam. A small positron emission tomography (PET) system was exposed in an ocular beam of the Proton Therapy Center at MD Anderson Cancer Center to view phantoms irradiated by 3.5 × 1010 protons with a kinetic energy of 75.8 MeV delivered in 101.5 ms-long spills yielding a dose rate of 164 Gy s−1. Most in-spill events were due to prompt gammas. Reconstructed post-spill tomographic events, recorded for up to 20 min, yielded quantitative imaging and dosimetric information. These findings open a new and novel modality for imaging and monitoring of FLASH proton therapy exploiting in-spill prompt gamma imaging followed by post-spill PET imaging.

Fascinating recent reports of preclinical cancer research indicate that beam radiation delivered with high doses and high dose rates-a FLASH beam-results in not-yet-understood and surprisingly effective eradication of tumors and unexpectedly promising sparing of healthy tissues (Favaudon et al 2014, Fouillade et al 2017, Montay-Gruel et al 2017, Patriarca et al 2018, Bourhis et al 2019, Montay-Gruel et al 2019, Vozenin et al 2019a, 2019b, Fouillade et al 2020, Hughes and Parsons 2020, Lin et al 2021).The underlying biomedical mechanism responsible for the FLASH effects must be understood and thoroughly elucidated if this emerging therapy modality is to be fully exploited and adopted.In fact, the discovery, or perhaps more accurately an overdue re-focus on these phenomena (Sutton and Rotblat 1957, Lajtha and Oliver 1961, Patricia and Lindop 1961, Prydz and Henriksen 1961, Randolph 1961, Smith 1962, Rotblat and Simmons 1963, Hall and Bedford 1964, Ellis 1968, Epp et al 1968, Hall and Berry 1968, Hall 1972, Bedford and Mitchell 1973, Mitchell et al 1979, Brenner et al 1991, Brenner and Hall 1991, Hall and David 1991), if confirmed in more extensive future studies, poses a profound and critical question of how well tumor eradication by radiation is really understood!The FLASH modality has the potential to revolutionize radiation oncology, thus it receives immense attention worldwide.FLASH-based treatments may impact the radiation therapy economy by substantially increasing the patient throughput, while improving the treatment outcome and decreasing postradiation side effects.
Ever since Wilson's seminal publication in 1946 (Wilson 1946), proton therapy has been recognized as one of the most promising techniques of radiation oncology (An informative website on particle).Over the years, it has benefitted from advances in medical accelerators, beam delivery, and the establishment of a comprehensive multi-step treatment planning process.This has been assisted and steadily refined by progress in modeling of beam interactions in phantoms and patients (Moreno et al 2019).On the other hand, monitoring and real-time assessment of each irradiation have not kept the same pace of improvements.The in vivo proton range verification-determination of the accuracy and efficacy of each irradiation-remains challenging today and is one of the hindering factors of progress.Better assessment of each irradiation is necessary, expected, and perhaps behind schedule in the era of significant advances in hardware technology, data processing, and image analysis.Creating an advanced protocol is not only a moral imperative but also an essential step to strengthen overall outcome of proton therapy.
A typical treatment comprises about 30 irradiation fractions (doses) of about 2 Gy each, delivered over a period of a few minutes once a day.Despite that positron-emitting isotopes are activated by a proton beam and positron emission tomography (PET) provides the most adequate mapping of tumor tissues, almost all routine clinical cases do not employ molecular imaging by PET to evaluate the therapy progression.This apparent shortcoming is due to the lack of proper in-beam PET scanners, excessive time necessary to revise a plan, and no sanctioned protocols compatible with administrative (insurance) allowances.Conventional injection-induced PET imaging of vulnerable patients undergoing the therapy is rare since common treatment-related inflammations obscure the clarity of the tumor images.
These limitations can be overcome by using optimized PET detectors, fast data processing, and utilizing a tissue activated by radiation (Parodi et al 2002, Enghardt et al 2004, Crespo et al 2006, 2007, Knopf et al 2009, Richard and Chevallier 2010, Knopf and Lomax 2013, Seco et al 2014, Verburg and Seco 2014, Seco and Spadea 2015, Jones et al 2018, Liu et al 2019, Shusharina et al 2019, Lang 2022).In fact, less conventional beam deliveries offer substantially stronger activation signal due to the increased beam intensity.A newer treatment plan type, not yet widely prescribed, uses a significantly reduced number of fractions ('hypofractionations'), depending on the tumor, may be on the order of 15-20 Gy (totaling to about 60 Gy).The emerging and promising, not-yet-clinically practiced plan, is FLASH therapy.This is a hypofractionated delivery technique featuring ultra-high dose (in the to-be-determined range, perhaps as high as 10-40 Gy) delivered in milliseconds or less, as opposed to conventional, several minutes-long beam delivered over a 20-30 d period.The FLASH idea is gaining prominence in view of reports indicating that FLASH radiotherapy results in a surprising, not yet understood, and unexpectedly promising effect of better sparing healthy tissues without compromising the eradication of tumors The instantaneous positron yield due to FLASH beam activation may be as much as 1,000 times higher than in conventionally fractionated proton beam.This poses constraints and challenges for instruments, including PET detectors surrounding the irradiated tissue, and presents opportunities for using the strong and fast signal emitted by isotopes activated by the proton beam, mostly 15 O, 13 N, or 11 C, and their minimal biological washout.It allows one to take a snapshot of tissues where the beam delivered its ionizing energy.The timing and the strength of this signal is unprecedented and unexploited.
Imaging and dosimetric information provided by a PET scanner during proton therapy can be employed for monitoring of irradiated tissues and adaptive revision of treatment plan after each radiation.This can be particularly effective with hypofractionations and FLASH fractionation.However, FLASH extraction is not available at many existing proton treatment centers and the necessary in-beam PET scanners are still being developed (Parodi et al 2002, Crespo et al 2006, 2007, Knopf et al 2009, Richard and Chevallier 2010, Knopf and Lomax 2013, Seco et al 2014, Verburg and Seco 2014, Seco and Spadea 2015, Jones et al 2018, Liu et al 2019, Shusharina et al 2019, Lang 2022).But even before then, the impact on medical personnel and instrumentation must be well characterized so that such fast extraction can be safely experimented with and ultimately be routinely employed in therapy.A PET scanner must function in the spatial and temporal proximity to the beam that creates an intense radiation zone, including penetrating (i.e.radiation damaging) low-energy neutrons, so it is critical and necessary to systematically and thoroughly explore the FLASH beam and its effects on the surroundings.
Our work presented here is directly motivated by great potential benefits of using in-beam PET imaging of FLASH proton beam.The unavoidable tissue activation by protons results in an untapped source of short-lived isotopes that can be used for imaging.We report here results of measurements of in-spill (i.e. during the beam extraction) and post-spill (i.e. after the beam extraction) gammas derived from activations produced by a FLASH proton beam of a sub-second duration time.Our recent limited-scope work (Abouzahr et al 2023) investigated a PMMA phantom exposed to the same FLASH beam.We were able to collect then only 3 min of data after the beam extraction (spill).
In the work here, we are reporting on beam exposures of six phantoms and analysis of data that were continuously recorded both during the 101.5 ms spill and up to 20 min after the spill, with no changes in data acquisition for the two periods.These are unmatched observations of a FLASH beam that allow kinetic imaging and dosimetry.These unprecedented studies pave a way to a new important PET modality that will improve the overall outcome of proton radiation therapy.

Methods
The experiment was carried out at the ocular beam of the Proton Therapy Center of the MD Anderson Cancer Center.This is a beam with FLASH extraction designed and characterized by the Center's staff (Titt et al 2021, Yang et al 2022).There were three main components of the experimental setup: the beam, the PET system, and phantoms.The PET system was calibrated using a linear 68 Ge calibration source (The Eckert-Ziegler).
Beam: A Hitachi medical synchrotron at MD Anderson delivers a 'pencil' beam that has a nominal energy of 87 MeV.This energy is reduced to 75.8 MeV after the beam passes through passive shaping elements and about 450 mm of air.The beam was delivered through a double-scattering system (Yang et al 2022) and a 30 mm diameter opening in a large transverse brass shield.The energy has a 2σ gaussian width of 1.16 MeV, and the intensity is flat to within 2%-3% across a disk of about 16 mm diameter, then it falls off to about 80% at 20 mm, about 50% at 25 mm, and about 5% at 30 mm diameter.A FLASH extraction (spill) lasts about 101.5 ms and PET system: Two LYSO scintillation crystal arrays (modules) were used to form a PET mini-scanner shown in figure 1.These modules were assembled earlier as spares for a larger PET scanner that was recently completed for preclinical tests in a proton beam therapy (TOF-PET for Proton Therapy (TPPT), Klein et al 2021, TPPT Consortium 2021).Each module comprises two 8 × 8 arrays assembled out of Lu 1.8 Y 0.2 SiO 5 :Ce (LYSO:Ce) scintillation crystals.Each 64-element LYSO array features 'pixel' crystals of dimensions 3.005 × 3.005 × 15 mm 3 (Tong) and each crystal is coupled to a Hamamatsu S14161-3050HS-08 silicon photomultiplier (SiPM), also known as a Multi-Pixel Photon Counter (MPPC) (Hamamatsu silicon photo-multipliers (SiPM)).All pixel crystals were polished and separated by three layers of an enhanced specular reflection (ESR) film (ESR) so that the pitch of crystals with wrappings closely matched the SiPM pixel pitch that features 8 × 8 pixels of 3.0 mm × 3.0 mm dimensions set 0.2 mm apart.
The two modules were placed in precisely fixed positions facing each other and equally distant on either side of phantoms that were centered on the beam axis.The central beam axis was contained in the horizontal symmetry plane of the mini-scanner and phantoms.Data were taken with three different distances of 110, 150, and 210 mm to the beam line.The field of view of this PET mini-scanner covered most of the phantom volumes where the beam energy was deposited.
The readout electronics was provided by PETsys Electronics (PETSys Electronics) and is identical to the readout employed in the Time-of-Flight for Proton Therapy (TPPT) PET scanner (TOF-PET for Proton Therapy (TPPT)).SiPMs were bumped-soldered to FEB/S boards integrated with readout electronics using the PETsys TOFPET2 ASIC in conjunction with the PETsys FEB/I ASIC interface board (PETSys Electronics, TOF-PET for Proton Therapy (TPPT), Klein et al 2021, TPPT Consortium 2021, Klein et al 2022).A Peltier cooling system was installed on top of LYSO arrays and front-end electronics to maintain stable SiPM and ASIC temperatures that was logged and monitored by the data acquisition system throughout each run.
Phantoms: Five different materials were used for making cylindrical phantoms.There were poly-methyl methacrylate (PMMA) cylindrical phantoms of 25.4 mm diameter and 101.6 mm length.The molecular composition of this phantom was ( ) C O H 5 2 8 n and a density of 1.18 g cm −3 .There were also phantoms made out of high-density polyethylene (HDPE) with the same dimensions, molecular composition of ( ) C H 2 4 n and a density of 0.94 g cm −3 .We also made distilled water phantoms (H 2 O), contained in a 37.6 mm diameter PMMA cylinder with 1.3 mm thick beam entry window and a 3.3 mm thick side wall.Two metal cylindrical phantoms of dimensions 63.5 mm diameter and 20.0 mm length were also irradiated.One was made out of 99.99% purity copper, and the second was formed out of an alloy composed mainly of 64.3% nickel (by mass), 32.0% copper, 2.0% of iron, and 1.2% of magnesium.

Results
The FLASH beam at the Proton Therapy Center of MD Anderson Cancer Center irradiated five different phantom types made out of PMMA, HDPE, water, copper, and nickel alloy (details are described in the 'Methods' section).Data were also collected with an 'empty target' (i.e. the air).The data logging, lasting from 3 to 20 min, was organized in runs associated with an individual FLASH extraction (spill).Most runs involved new (i.e.never exposed to beam) phantoms.In each run, the data acquisition (DAQ) commenced a few tens of seconds before a FLASH spill that lasted 101.5 ms.This early start was necessary to assure that the DAQ system was ready for a fast extraction (i.e. a FLASH spill) which required some preparatory actions by the accelerator staff.After each spill, data were collected until the end of a preset time interval.Table 1 summarizes information about all 13 runs during which we acquired data.
We have recently reported on the first data registration and PET imaging in this beam (Abouzahr et al 2023).That preliminary work introduced us to the radiation environment of this beam although our instruments did not record any in-spill data then.Due to the Proton Therapy Center's regular operations, both times the access to the proton FLASH beam was limited.We started by taking three 3min-long runs with PET modules placed 210 cm, the largest possible distance away from the beam's center line, which was also the center line of all inbeam phantoms.The first run was taken with 40 mm thick copper shielding plates in front of PET modules; the second run had 20 mm thick copper shielding; and during the third run PET modules were unshielded.Each run was immediately analyzed and evaluated with respect to a possible in-spill and post-spill dead-time.We judged, as discussed below, that later runs can be taken without the copper shielding.
However, high data rates during 101.5 ms spills resulted in saturation of in-spill data.We also observed that during some post-spill periods, occasional temporal data acquisition dead times were manifested by missing data packets.These dead-times did not affect the assessment of data.Although the three initial short and exploratory Runs 1, 2, and 3, which tested whether the PET modules required shielding, contributed to our general FLASH beam studies, we report here on runs taken with unshielded PET modules and pristine unactivated phantoms.Run 4 was taken with a new PMMA phantom #2, Run 5 was with a new HDPE phantom #1, and Runs 6 and 7 were taken with new PMMA phantoms #3 and #4 with PET distances to the beam center of 150 and 110 mm, respectively.Run 8 used a water phantom (water contained in a PMMA tube) and Run 9  with an 'empty target' (i.e.air).Both of these runs showed diffusion of positron-emitting isotopes, which in the case of air led to a quick dissipation of signal (as demonstrated by a low number of post-spill coincidences which was not due only to the low density of air).Runs 10 and 13 were conducted as a systematic check of the same earlier configurations.Runs 11 and 12 used compact metallic phantoms to connect measurements to a possible future enhancement of imaging and/or theranostics with gold nanoparticles (Smith et  Figure 2 shows Run 4 results and illustrates a typical time spectrum of the number of coincidences between the two PET modules reconstructed for an entire run.Panel (a) shows an overall time structure of recorded coincidences that features a low pre-spill (i.e.pre-beam extraction) activity associated with long-lived isotopes from some earlier irradiations (see table 1), a very high activity during the spill (the 'in-spill' spike), and a postspill activity evolving according to half-lives of activated isotopes.Panel (b) zooms in on the non-uniform inspill beam intensity measured by the accelerator and beam line's monitoring chamber during the 101.5mslong beam extraction.We also show there the intensity measured by the PET modules, as described in the 'In-spill events' section below.Three panels in figure 3 show the SiPM charge spectra recorded by the front-end electronics (this charge is proportional to the energy deposited in a LYSO crystal) for three time intervals: full run, in-spill, and post-spill.These spectra demonstrate that post-spill events exhibit a clear photopeak and thus are due to positron-emitting species (PES), while the in-spill charge spectrum is derived from random coincidences of two prompt gammas.
In-spill events During FLASH spills lasting 101.5 ms, we were able to record about 8.1 million single events and reached the front-end electronics saturation limit.Since the beam saturated the readout and we did not have any collimators for gammas, we were unable to use these events for imaging, e.g. by SPECT reconstruction (Verburg et  For events in two similar detectors, randomly and independently occurring in time, it is expected from general statistical considerations that the rate of random coincidences is proportional to the product of rates in each detector.In our case, by squaring the coincidence rates, we observe (see figure 2(b)) a very strong correlation with the beam intensity measured by independent beam monitors.This conclusion was tested and unchanged by varying the 500 ps coincidence time window from 100 ps to 100 ns.We conclude that by using inspill random prompt gammas, we are able to measure the beam intensity and thus can relate it to the delivered dose.Despite the DAQ saturation, a squared rate of coincidences formed by random events in the two PET modules (essentially an identical requirement to the PET reconstruction) tracks the beam intensity very well.We conclude that further use of in-spill prompt gammas requires a collimator that would lower the rate and, with appropriate support of simulations, would allow SPECT-like imaging and dosimetry.In our simulations of the experiment we focused on the post-spill activity due to the DAQ saturation effect during the spill.
Post-spill events Post-spill events collected by two-module PET data acquisition for up to 20 min are essentially all due to 511 keV gammas emitted by β + decays of isotopes activated by the beam.Each event was associated with one PET LYSO pixel crystal on the left of the beam and one in a module on the right of the beam and formed a line of response (LOR).These lines of response are used in further analyses that included their timing since spills, imaging, and dosimetry.The number of LORs as a function of time are plotted in the left panel of figure 2 and in figure 4.
Figure 4 shows time distribution for post-spill events of selected phantoms and fits of time-dependent activities by main positron-emitting isotopes.Histograms also include results of simulations.We simulated proton beam interactions with the phantoms using the Geant4 toolkit (Geant4 2016, Abouzahr et al 2023).The QGSP_BIC_ HP reference physics list was employed, enabling high precision neutron and electromagnetic physics modeling.To obtain the beam-induced activity for the metallic phantoms Geant4 was used directly, while for the PMMA, HDPE and water phantoms we have applied a custom generator for positron-emitting species using the experimentally-validated production cross-sections from (Bauer et al 2013) and the tracking information for the primary protons from Geant4.Known uncertainties in modeling stem from an incomplete list of activated isotopes and also include the lack of knowledge of activation cross-sections in the specific energy range of the experiment.However, the overall agreement with post-spill measurements is remarkably good.
The relative abundance of the dominating positron-emitting isotopes activated by the beam spill in each phantom was determined by fitting each post-spill activity spectrum shown in figure 4 by a sum of activities of selected isotopes.The chemical composition of each phantom (as described in the 'Methods' section) and literature reports (Akagi et al 2013, Kraan et al 2014) guided the selection of isotopes used in fits.Additionally, for

Note.
a We note that the water phantom suffers from the diffusion process and thus the effective activity accounts for a mix of activated and un-activated water which leads to the poor fit on figure 4. Furthermore, we note that the diffusion time in water is faster than PET data acquisition.The uncertainty for all the activity values in the table are within 10%.
the metal phantoms, the majority of radioisotopes activated in copper and nickel have half-lives that are on the order of hours or days (Garrido et al 2016) and thus much longer that our data acquisition times.As a result we selected only a few isotopes indicated in figure 4. All fits are summarized in table 2.
The exponential nuclear decay functions use a least-squares algorithm provided by Minuit, a numerical minimization software (James and Roos 1975).For each run, the half-life in the decay functions were assumed The corresponding exponential decay fits for each distribution are also plotted.In the inset of each subplot, a schematic of the experimental setup is displayed with the relevant spatial region highlighted in grey on each module (note that the distance between the PET modules and the phantom is not to scale).The grey arrows illustrate the beam direction.The distributions by panel go as follows: (a) PMMA from 0.0 to 12.6 mm, (b) HDPE from 0.0 to 12.6 mm, (c) PMMA from 12.8 to 25.4 mm, (d) HDPE from 12.8 to 25.4 mm, (e) PMMA from 26.2 to 38.9 mm, (f) HDPE from 26.2 to 38.9 mm, (g) PMMA from 39.1 to 51.7 mm, and (h) HDPE from 39.1 to 51.7 mm.based on the known half-lives of the expected, dominating isotopes for each phantom as listed in table 2. The result of fitting produced the corresponding coefficients for each isotope yielding information about their relative activity levels.For instance, the fitting of activity function ( )  t associated with the PMMA phantoms was given by ( ) , where the amplitudes A, B, C, and D are the fit parameters and expressions • ( ) t Y ln 2 x denote the half-lives of isotopes xY .Finally, the constant term E accounts for a constant (during a run) environmental background and/or due to activation in earlier spills if a phantom was irradiated more than once.This parameter is determined by fitting to the data before a spill.Table 2 summarizes the isotopic abundance as determined by these fit parameters and from Geant4 simulations.
Due to the energy dependence of positron-emitting isotopes, their abundance is depth related, as indicated in figure 17 or our earlier publication (Abouzahr et al 2023).We consider this effect by subdividing the field of view into four spatial vertical regions composed of four columns of detector pixels, each probing approximately 12.6 mm sections along the phantoms (see the bottom illustration in figure 1).Further division limits the statistical power of the measurements.We then separately fit the PET data from each spatial region to the exponential nuclear decay functions, following the same procedure used for figure 4. Figure 5 displays the time distributions for the post-spill coincidences for the 210 mm PMMA and HDPE data with their respective decay function fits at the four depth ranges.Table 3 shows the relative isotopic abundance at each depth range in the PMMA and HDPE phantoms calculated from the fit parameters of figure 5.The results clearly show that the production of the positron-emitting isotopes strongly depend on depth and, hence, beam energy.Further analysis of these data is underway and includes detailed comparison with simulations.
For any given activated isotope in a phantom, its abundance is calculated as the ratio of its activity (i.e. its fit parameter) to the sum of all activities (i.e. the sum of all fit parameters).Despite expected contributions from 13 N in the PMMA and water phantoms, the fits did not yield any significant 13 N activity.We also note that the water phantom suffers from the diffusion process and thus the effective activity accounts for a mix of activated and un-  activated water.For copper and nickel alloy we report fits with and without 61 Zn and can only compare the total yield.

Imaging-time evolution and dosimetry
We have used the CASToR imaging package (Merlin et al 2018) to create images of positron-emitting volumes by using LORs recorded during the runs.Reconstruction of events can be done iteratively many times but in our case the tomographic (i.e.angular) field of view is very limited so an overall good quality contrast image is usually obtained with a relatively low number of iterations.Upon examining up to 50 iterations, we selected the 5th image iteration as qualitatively the sharpest and best reflecting the beam-phantom configuration.The two canonical views, coronal and transaxial, are shown in figure 6.In addition to images we show the dosimetric counts in PET crystals represented by the black bars which reflect the number of PET coincidences detected in columns (coronal view) and rows (transaxial view) of the crystal arrays.Red bars denote columns and rows that include either a dead or a noisy pixel.These figures constitute the synthesis of our imaging and dosimetry measurements.
We note that the lines of response of PET modules were effectively almost perpendicular to the direction of the beam and the field of view of the PET is depicted by straight lines connecting the modules.The images and dosimetric histograms clearly show the rise of registered activity with the beam penetration, as expected from the simulations shown in figure 4 and illustrate the power of the measurements.
The time duration of runs and the statistics of PET events enables us to conduct a test of kinetics of the PET signal as a function of time.Figure 7 shows frames of coronal PET imaging captured in 120 s or 90 s time intervals for PMMA, HDPE, and water phantoms, respectively.These snapshots come from an animated kinetic reconstruction of images.We note that images formed from beam-activated isotopes generally reflect the position of the Bragg peak but are convolved with cross-sections for producing β + emitters.Additionally, the water phantom exhibits fast diffusion of activated isotopes which also mix with unactivated water.It should be stressed that the test was conducted with just two PET modules so a cylindrical scanner with many such modules would have much larger event statistics.This would substantially improve the quality of imaging and the time flow of images and dosimetry could be much shorter and of higher resolution.

Discussion
We have conducted pioneering measurements of in-spill and post-spill gamma emissions in a FLASH proton beam and thus establishing the feasibility of new imaging modalities.This work demonstrates the unprecedented proof-of-principle for the capabilities of an in-beam PET scanner for imaging and dosimetry of a FLASH proton beam.FLASH extraction naturally divides data into in-spill events due to prompt gammas and post-spill events due to positron-emission.The in-spill data present a novel 'snapshot' source of events for determining the proton range.Imaging based on these data is free of motion blur or biological washout but requires a new in-beam detector (perhaps with a collimator).The post-spill data allows one to monitor the evolution of the activity of irradiated tissues and a related washout effect.Both techniques combined provide untried feedback on proton range and dosimetry.These first trials and results of FLASH data analysis based on a minimal, unoptimized hardware setup deliver encouraging news for PET imaging of FLASH beams and allows one to expect much more information from a larger, suitably designed detector.
Results presented here open a new PET modality with proton beams which is particularly attractive for FLASH therapy but can serve effectively all proton irradiations thus leading to improved monitoring of irradiations and image-guided proton therapy.A conventional therapy with about 30 irradiations make it difficult for mid-course adjustments due to current treatment protocols and administrative therapy constraints.A new, FLASH-based therapy could be designed from the very beginning as an image-guided irradiation plan and thus not only improve overall therapy outcomes but also increase the patient reach and throughput.

Figure 1 .
Figure 1.Experimental setup.(a) A PET module comprised of 2 × 64 scintillating LYSO crystal arrays as schematically shown in the inset.(b) A photograph of a complete setup and a PMMA phantom installed in the beam.(c) and (d) Schematic views from above and looking upstream, respectively, of a complete setup at one of the exposure configurations.The beam direction is indicated by an orange arrow.

Figure 2 .
Figure 2. (a) The number of PET coincidences as a function of time across a 20 min data acquisition period for PMMA 2 at a separation distance of 210 mm.The spill occurred at approximately the 62 s mark.The activity shown before the spill arises from the background radiation environment induced from prior spills.(b) A comparison between PET random single coincidences to the beam monitor output voltage per 1 ms.The mean normalized ratio between the two quantities as a function of time is displayed in the inset.

Figure 3 .
Figure 3.A typical energy spectrum in a single crystal pixel collected during the entire run [Panel (a)], during the beam spill [Panel (b)], and after the beam spill [Panel (c)].The photopeak at DAQ counts of about 27 is due to 511 keV gammas.The lack of a photopeak in the energy spectra for the data collected during the beam extraction (the 'in-spill' data) evinces that the radiation environment during the beam spill is dominated by prompt gammas.

Figure 4 .
Figure4.Post-spill PET coincidence data and fits as a function of time for phantom runs with the PET modules placed at 210 mm from the beam center line.Data are shown as a filled blue histogram, fits are marked by solid lines, while results of Geant4 simulations are shown with dashed lines.The relative activity (i.e. the abundance of the positron-emitting isotopes produced in these phantoms, as determined by these fits) is summarized in table 2.
al 2012, Polf et al 2013, Verburg and Seco 2014, Werner et al 2017, Hueso-González et al 2018, Aleksandra Wrońska and for the SiFi-CC group 2020, Tattenberg et al 2023).However, we processed the in-spill data by counting the rate of coincidences of single hits in the two PET modules over a 500 ps time window.

Figure 5 .
Figure5.The post-spill coincidence time distributions from the 210 mm PMMA and HDPE data at different depths along each phantom.The corresponding exponential decay fits for each distribution are also plotted.In the inset of each subplot, a schematic of the experimental setup is displayed with the relevant spatial region highlighted in grey on each module (note that the distance between the PET modules and the phantom is not to scale).The grey arrows illustrate the beam direction.The distributions by panel go as follows: (a) PMMA from 0.0 to 12.6 mm, (b) HDPE from 0.0 to 12.6 mm, (c) PMMA from 12.8 to 25.4 mm, (d) HDPE from 12.8 to 25.4 mm, (e) PMMA from 26.2 to 38.9 mm, (f) HDPE from 26.2 to 38.9 mm, (g) PMMA from 39.1 to 51.7 mm, and (h) HDPE from 39.1 to 51.7 mm.

Figure 6 .
Figure 6.The coronal (top panel) and axial (bottom panel) views of PET images produced using the CASToR imaging package (Merlin et al 2018).The color scale is arbitrary.The cylindrical PMMA 2 phantom is outlined by a rectangle and a circle, and the field of view boundaries are indicated by orange lines.Data were taken in Run 4. For creating these images we used a voxel size of 1.5 × 1.5 × 1.5 mm 3 and the results of iteration 5 are shown.The black bars denote the sum of coincidence counts in columns or rows of pixels.Red bars signify a column or a row that includes either a dead or a noisy pixel and thus the counts are less reliable.

Figure 7 .
Figure 7. Frames of coronal PET imaging captured in 120 s long or 90 s longtime periods, marked above each frame, for PMMA, HDPE, and water phantoms, respectively.As in figure 6, the field of view boundaries are indicated by orange lines.The Geant4 simulated activity produced by the proton beam is shown in two-dimensional plot on the right.As commented in the text, the activation cross-sections and diffusion in water affect the imaging of this phantom.

Table 1 .
Summary of run configurations, PET module distances to the beam line, and statistics of 13 FLASH extractions (spills).

Table 2 .
The abundance of activated positron-emitting isotopes for each phantom is derived from the post-spill fits shown in figure 4.

Table 3 .
The abundance of activated positron-emitting isotopes for the PMMA and HDPE phantoms as a function of depth derived from the postspill fits shown in figure5.The uncertainty for all the activity values in the table are within about 15%.