Prediction of positron emitter distributions for range monitoring in carbon ion therapy: an analytical approach

An inception network for positron emission tomography based dose estimation in carbon ion therapy

Objective. We aim to evaluate a method for estimating 1D physical dose deposition profiles in carbon ion therapy via analysis of dynamic PET images using a deep residual learning convolutional neural network (CNN). The method is validated using Monte Carlo simulations of 12C ion spread-out Bragg peak (SOBP) profiles, and demonstrated with an experimental PET image. Approach. A set of dose deposition and positron annihilation profiles for monoenergetic 12C ion pencil beams in PMMA are first generated using Monte Carlo simulations. From these, a set of random polyenergetic dose and positron annihilation profiles are synthesised and used to train the CNN. Performance is evaluated by generating a second set of simulated 12C ion SOBP profiles (one 116 mm SOBP profile and ten 60 mm SOBP profiles), and using the trained neural network to estimate the dose profile deposited by each beam and the position of the distal edge of the SOBP. Next, the same methods are used to evaluate the network using an experimental PET image, obtained after irradiating a PMMA phantom with a 12C ion beam at QST’s Heavy Ion Medical Accelerator in Chiba facility in Chiba, Japan. The performance of the CNN is compared to that of a recently published iterative technique using the same simulated and experimental 12C SOBP profiles. Main results. The CNN estimated the simulated dose profiles with a mean relative error (MRE) of 0.7% ± 1.0% and the distal edge position with an accuracy of 0.1 mm ± 0.2 mm, and estimate the dose delivered by the experimental 12C ion beam with a MRE of 3.7%, and the distal edge with an accuracy of 1.7 mm. Significance. The CNN was able to produce estimates of the dose distribution with comparable or improved accuracy and computational efficiency compared to the iterative method and other similar PET-based direct dose quantification techniques.

Energy spread estimation of radioactive oxygen ion beams using optical imaging

Radioactive ion (RI) beams combined with in-beam positron emission tomography enable accurate in situ beam range verification in heavy ion therapy. However, the energy spread of the radioactive beams generated as secondary beams is wider than that of conventional stable heavy ion beams which causes Bragg peak region and distal falloff region broadening. Therefore, the energy spread of the RI beams should be measured carefully for their quality control. Here, we proposed an optical imaging technique for the energy spread estimation of radioactive oxygen ion beams. A polymethyl methacrylate phantom (10.0 × 10.0 × 9.9 cm3) was irradiated with an 15O beam (mean energy = 247.7 MeV u−1, standard deviation = 6.8 MeV u−1) in the Heavy Ion Medical Accelerator in Chiba. Three different momentum acceptances of 1%, 2% and 4% were used to get energy spreads of 1.9 MeV u−1, 3.4 MeV u−1 and 5.5 MeV u−1, respectively. The in-beam luminescence light and offline beam Cerenkov light images were acquired with an optical system consisting of a lens and a cooled charge-coupled device camera. To estimate the energy spread of the 15O ion beams, we proposed three optical parameters: (1) distal-50% falloff length of the prompt luminescence signals; (2) full-width at half maximum of the Cerenkov light signals in the beam direction; and (3) positional difference between the peaks of the Cerenkov light and the luminescence signals. These parameters estimated the energy spread with the respective mean squared errors of 2.52 × 10−3 MeV u−1, 5.91 × 10−3 MeV u−1, and 0.182 MeV u−1. The distal-50% falloff length of the luminescence signals provided the lowest mean squared error among the optical parameters. From the findings, we concluded optical imaging using luminescence and Cerenkov light signals offers an accurate energy spread estimation of 15O ion beams. In the future, the proposed optical parameters will be used for energy spread estimation of other RI beams as well as stable ion beams.

Use of Gaussian-type functions for flux-based dose calculations in carbon ion therapy

In radiation therapy, it is very important to ensure that the radiation dose is correctly delivered to the patient. This is achieved by obtaining quantitative dose measurements for beam calibration in the treatment planning system. Dose calculations should be performed with the required accuracy to a degree of uncertainty of less than 1%. The measurement of the absorbed dose in and around body tissues irradiated with carbon ions requires careful use of materials selected from established phantom and radiation detectors. The main advantage of such materials is that when information on the energy and nature of charged particles at the desired point is incomplete or inaccurate, they can allow determination of the absorbed dose. In general, radiation interactions in a tissue representation caused by carbon ions can be characterized by calculating the linear stopping power. Carbon ions have a limited penetration depth within human tissues that depends on the energy and stopping power of these ions as they penetrate into the body. The purpose of the present study was to calculate the stopping power, range and dose to intestinal and prostate tissues of carbon ions. The stopping power values of these tissues were specified by the effective charge approach method. The 5ZaPa-NR-CV, pcemd-4 and pcSseg-4 sets of Gaussian-type functions were employed for the calculation of electronic charge density. Range calculations were made by means of the Gaussian quadrature method, making use of the continuous slowing down approximation. Flux-based dose calculations were also carried out in accordance with the Bragg-Gray theorem using the Geant4 and FLUKA simulation toolkits. The results were compared with each other and with the SRIM and CasP datasets. Then, depth-dose distributions and range values were verified by positron emission activity using the GATE toolkit. Among the different types of Gaussian functions used here, the best semi-analytical result was found for the 5ZaPa-NR-CV set. The results obtained in the present study can be used for dose verification and dose reconstruction in charged particle radiotherapy and for radiation research on the interaction of radiation with matter. The results calculated here will be useful for quantifying uncertainties associated with stopping power, range, and reconstruction of dose in charged particle therapy.

A filtering approach for PET and PG predictions in a proton treatment planning system

Positron emission tomography (PET) and prompt gamma (PG) detection are promising proton therapy monitoring modalities. Fast calculation of the expected distributions is desirable for comparison to measurements and to develop/train algorithms for automatic treatment error detection. A filtering formalism was used for positron-emitter predictions and adapted to allow for its use for the beamline of any proton therapy centre. A novel approach based on a filtering formalism was developed for the prediction of energy-resolved PG distributions for arbitrary tissues. The method estimates PG yields and their energy spectra in the entire treatment field. Both approaches were implemented in a research version of the RayStation treatment planning system. The method was validated against PET monitoring data and Monte Carlo simulations for four patients treated with scanned proton beams. Longitudinal shifts between profiles from analytical and Monte Carlo calculations were within -1.7 and 0.9 mm, with maximum standard deviation of 0.9 mm and 1.1 mm, for positron-emitters and PG shifts, respectively. Normalized mean absolute errors were within 1.2 and 5.3%. When comparing measured and predicted PET data, the same more complex case yielded an average shift of 3 mm, while all other cases were below absolute average shifts of 1.1 mm. Normalized mean absolute errors were below 7.2% for all cases. A novel solution to predict positron-emitter and PG distributions in a treatment planning system is proposed, enabling calculation times of only a few seconds to minutes for entire patient cases, which is suitable for integration in daily clinical routine.