Machine learning approach for proton range verification using real-time prompt gamma imaging with Compton cameras: addressing the total deposited energy information gap

Objective.Compton camera imaging shows promise as a range verification technique in proton therapy. This work aims to assess the performance of a machine learning model in Compton camera imaging for proton beam range verification improvement.Approach.The presented approach was used to recognize Compton events and estimate more accurately the prompt gamma (PG) energy in the Compton camera to reconstruct the PGs emission profile during proton therapy. This work reports the results obtained from the Geant4 simulation for a proton beam impinging on a polymethyl methacrylate (PMMA) target. To validate the versatility of such an approach, the produced PG emissions interact with a scintillating fiber-based Compton camera.Main results.A trained multilayer perceptron (MLP) neural network shows that it was possible to achieve a notable three-fold increase in the signal-to-total ratio. Furthermore, after event selection by the trained MLP, the loss of full-energy PGs was compensated by means of fitting an MLP energy regression model to the available data from true Compton (signal) events, predicting more precisely the total deposited energy for Compton events with incomplete energy deposition.Significance.A considerable improvement in the Compton camera's performance was demonstrated in determining the distal falloff and identifying a few millimeters of target displacements. This approach has shown great potential for enhancing online proton range monitoring with Compton cameras in future clinical applications.

Measurement of the time structure of FLASH beams using prompt gamma rays and secondary neutrons as surrogates

OBJECTIVE

The aim of this study was to investigate the feasibility of online monitoring of irradiation time (IRT) and scan time for FLASH proton radiotherapy using a pixelated semiconductor detector.
Approach: Measurements of the time structure of FLASH irradiations were performed using fast, pixelated spectral detectors based on the Timepix3 (TPX3) chips with two architectures: AdvaPIX-TPX3 and Minipix-TPX3. The latter has a fraction of its sensor coated with a material to increase sensitivity to neutrons. With little or no dead time and an ability to resolve events that are closely spaced in time (tens of nanoseconds), both detectors can accurately determine IRTs as long as pulse pile-up is avoided. To avoid pulse pile-up, the detectors were placed well beyond the Bragg peak or at a large scattering angle. Prompt gamma rays and secondary neutrons were registered in the detectors' sensors and IRTs were calculated based on timestamps of the first charge carriers (beam-on) and the last charge carriers (beam-off). In addition, scan times in x, y, and diagonal directions were measured. The experiment was carried out for various setups: i) a single spot, ii) a small animal field, iii) a patient field, and iv) an experiment using an anthropomorphic phantom to demonstrate in vivo online monitoring of IRT. All measurements were compared to vendor log files. 
Main results: Differences between measurements and log files for a single spot, a small animal field, and a patient field were within 1%, 0.3% and 1%, respectively. In vivo monitoring of IRTs (95-270 ms) was accurate within 0.1% for AdvaPIX-TPX3 and within 6.1% for Minipix-TPX3. The scan times in x, y, and diagonal directions were 4.0, 3.4, and 4.0 ms, respectively. 
Significance: Overall, the AdvaPIX-TPX3 can measure FLASH IRTs within 1% accuracy, indicating that prompt gamma rays are a good surrogate for primary protons. The Minipix-TPX3 showed a somewhat higher discrepancy, likely due to the late arrival of thermal neutrons to the detector sensor and lower readout speed. The scan times (3.4 ± 0.05 ms) in the 60-mm distance of y-direction were slightly less than (4.0 ± 0.06 ms) in the 24-mm distance of x-direction, confirming the much faster scanning speed of the Y magnets than that of X. Diagonal scan speed was limited by the slower X magnets.

TOPAS-imaging: extensions to the TOPAS simulation toolkit for medical imaging systems

Objective. The TOol for PArticle Simulation (TOPAS) is a Geant4-based Monte Carlo software application that has been used for both research and clinical studies in medical physics. So far, most users of TOPAS have focused on radiotherapy-related studies, such as modeling radiation therapy delivery systems or patient dose calculation. Here, we present the first set of TOPAS extensions to make it easier for TOPAS users to model medical imaging systems.Approach. We used the extension system of TOPAS to implement pre-built, user-configurable geometry components such as detectors (e.g. flat-panel and multi-planar detectors) for various imaging modalities and pre-built, user-configurable scorers for medical imaging systems (e.g. digitizer chain).Main results. We developed a flexible set of extensions that can be adapted to solve research questions for a variety of imaging modalities. We then utilized these extensions to model specific examples of cone-beam CT (CBCT), positron emission tomography (PET), and prompt gamma (PG) systems. The first of these new geometry components, the FlatImager, was used to model example CBCT and PG systems. Detected signals were accumulated in each detector pixel to obtain the intensity of x-rays penetrating objects or prompt gammas from proton-nuclear interaction. The second of these new geometry components, the RingImager, was used to model an example PET system. Positron-electron annihilation signals were recorded in crystals of the RingImager and coincidences were detected. The simulated data were processed using corresponding post-processing algorithms for each modality and obtained results in good agreement with the expected true signals or experimental measurement.Significance. The newly developed extension is a first step to making it easier for TOPAS users to build and simulate medical imaging systems. Together with existing TOPAS tools, this extension can help integrate medical imaging systems with radiotherapy simulations for image-guided radiotherapy.

Applications of Machine Learning to Improve the Clinical Viability of Compton Camera Based in vivo Range Verification in Proton Radiotherapy

We studied the application of a deep, fully connected Neural Network (NN) to process prompt gamma (PG) data measured by a Compton camera (CC) during the delivery of clinical proton radiotherapy beams. The network identifies 1) recorded "bad" PG events arising from background noise during the measurement, and 2) the correct ordering of PG interactions in the CC to help improve the fidelity of "good" data used for image reconstruction. PG emission from a tissue-equivalent target during irradiation with a 150 MeV proton beam delivered at clinical dose rates was measured with a prototype CC. Images were reconstructed from both the raw measured data and the measured data that was further processed with a neural network (NN) trained to identify "good" and "bad" PG events and predict the ordering of individual interactions within the good PG events. We determine if NN processing of the CC data could improve the reconstructed PG images to a level in which they could provide clinically useful information about the in vivo range and range shifts of the proton beams delivered at full clinical dose rates. Results showed that a deep, fully connected NN improved the achievable contrast to noise ratio (CNR) in our images by more than a factor of 8x. This allowed the path, range, and lateral width of the clinical proton beam within a tissue equivalent target to easily be identified from the PG images, even at the highest dose rates of a 150 MeV proton beam used for clinical treatments. On average, shifts in the beam range as small as 3 mm could be identified. However, when limited by the amount of PG data measured with our prototype CC during the delivery of a single proton pencil beam (~1 × 10⁹ protons), the uncertainty in the reconstructed PG images limited the identification of range shift to ~5 mm. Substantial improvements in CC images were obtained during clinical beam delivery through NN pre-processing of the measured PG data. We believe this shows the potential of NNs to help improve and push CC-based PG imaging toward eventual clinical application for proton RT treatment delivery verification.

Characterization of two extraordinary AmLi neutron sources

Two legacy AmLi sources have been actively used across several projects within Pacific Northwest National Laboratory, however, applicability of source measurement data has been substantially limited due to the lack of original manufacturer documentation and cross comparison to other sources. This paper summarizes measurements conducted to better quantify source radionuclide content, neutron energy spectrum and emission rate of the AmLi sources. The measured neutron energy spectra of AmLi sources produced by Monsanto Research Corporation have not been found in open literature before and therefore the focus of the study was to obtain the high definition energy distribution of the sources. The initial gamma spectrometry measurements, aimed to estimate the radionuclide content and potential impurities, revealed a significant contamination of the source material with beryllium. Consequent neutron spectrum measurements confirmed beryllium contamination through the presence of a high energy tail contributing up to 17% to the total neutron emission rate. The obtained lithium-only portion of the measured source energy distribution was compared against other AmLi spectra discussed in the past and recent literature. Determination of the total neutron yields of the sources was made with measurements in a neutron well counter. Detection efficiency of the well counter was calculated specifically for these sources via detailed Monte Carlo simulations based on the obtained neutron energy distribution.

A simulation study investigating a Cherenkov material for use with the prompt gamma range verification in proton therapy

In vivo range verification methods will reveal information about the penetration depth into a patient for an incident proton beam. The prompt gamma (PG) method is a promising in vivo technique that has been shown to yield this range information by measuring the escaping MeV photons given a suitable detector system. The majority of current simulations investigating PG detectors utilize common scintillating materials ideal for photons within a low neutron background radiation field using complex geometries or novel designs. In this work we simulate a minimal detector system using a material ideal for MeV photon detection in the presence of a significant neutron field based on the Cherenkov phenomenon. The response of this selected material was quantified for the escaping particles commonly found in proton therapy applications and the feasibility of using the PG technique for this detector material was studied. Our simulations found that the majority of the range information can be determined by detecting photons emitted with a timing window less than ∼50 ns after the interaction of the proton beam with the water phantom and with an energy threshold focusing on the energy range of the de-excitation of 16O photons (∼6 MeV). The Cherenkov material investigated is able to collect these photons and estimate the range with timescales on the order of tens of nanoseconds as well as greatly suppress the signal due to neutron.

Range Verification Methods in Particle Therapy: Underlying Physics and Monte Carlo Modeling

Hadron therapy allows for highly conformal dose distributions and better sparing of organs-at-risk, thanks to the characteristic dose deposition as function of depth. However, the quality of hadron therapy treatments is closely connected with the ability to predict and achieve a given beam range in the patient. Currently, uncertainties in particle range lead to the employment of safety margins, at the expense of treatment quality. Much research in particle therapy is therefore aimed at developing methods to verify the particle range in patients. Non-invasive in vivo monitoring of the particle range can be performed by detecting secondary radiation, emitted from the patient as a result of nuclear interactions of charged hadrons with tissue, including β (+) emitters, prompt photons, and charged fragments. The correctness of the dose delivery can be verified by comparing measured and pre-calculated distributions of the secondary particles. The reliability of Monte Carlo (MC) predictions is a key issue. Correctly modeling the production of secondaries is a non-trivial task, because it involves nuclear physics interactions at energies, where no rigorous theories exist to describe them. The goal of this review is to provide a comprehensive overview of various aspects in modeling the physics processes for range verification with secondary particles produced in proton, carbon, and heavier ion irradiation. We discuss electromagnetic and nuclear interactions of charged hadrons in matter, which is followed by a summary of some widely used MC codes in hadron therapy. Then, we describe selected examples of how these codes have been validated and used in three range verification techniques: PET, prompt gamma, and charged particle detection. We include research studies and clinically applied methods. For each of the techniques, we point out advantages and disadvantages, as well as clinical challenges still to be addressed, focusing on MC simulation aspects.

Detecting prompt gamma emission during proton therapy: the effects of detector size and distance from the patient

Recent studies have suggested that the characteristics of prompt gammas (PGs) emitted from excited nuclei during proton therapy are advantageous for determining beam range during treatment delivery. Since PGs are only emitted while the beam is on, the feasibility of using PGs for online treatment verification depends greatly on the design of highly efficient detectors. The purpose of this work is to characterize how PG detection changes as a function of distance from the patient as a means of guiding the design and usage of clinical PG imaging detectors. Using a Monte Carlo model (GEANT4.9.4) we studied the detection rate (PGs per incident proton) of a high purity germanium detector for both the total PG emission and the characteristic 6.13 MeV PG emission from (16)O emitted during proton irradiation. The PG detection rate was calculated as a function of distance from the isocenter of the proton treatment nozzle for: (1) a water phantom irradiated with a proton pencil beam and (2) a prostate patient irradiated with a scanning beam proton therapy treatment field (lateral field size: ∼6 cm × 6 cm, beam range: 23.5 cm). An analytical expression of the PG detection rate as a function of distance from isocenter, detector size, and proton beam energy was then developed. The detection rates were found to be 1.3 × 10(-6) for oxygen and 3.9 × 10(-4) for the total PG emission, respectively, with the detector placed 11 cm from isocenter for a 40 MeV pencil beam irradiating a water phantom. The total PG detection rate increased by ∼85 ± 3% for beam energies greater than 150 MeV. The detection rate was found to be approximately 2.1 × 10(-6) and 1.7 × 10(-3) for oxygen and total PG emission, respectively, during delivery of a single pencil beam during a scanning beam treatment for prostate cancer. The PG detection rate as a function of distance from isocenter during irradiation of a water phantom with a single proton pencil beam was described well by the model of a point source irradiating a cylindrical detector of a known diameter over the range of beam energies commonly used for proton therapy. For the patient studies, it was necessary to divide the point source equation by an exponential factor in order to correctly predict the falloff of the PG detection rate as a function of distance from isocenter.