Noise evaluation of prompt-gamma technique for proton-therapy range verification using a Compton Camera

Real-time monitoring techniques are receiving increased attention due to their great potential for improving the precision of treatment delivery for hadron-therapy (HT). The study of prompt-γ(PG) focuses on the energetic photon radiation exiting the patient few nanoseconds after the beam irradiation. Its high intensity compared to other similar technologies like in-beam Positron Emission Tomography (PET) and good correlation with the original dose deposition profile are promising features. On the other hand, the use of neutral particles to monitor the dose comes together with a high noise background, mainly due to neutrons and scattered γ, which are uncorrelated with the original dose map. The search of a robust detector that fully exploits the information from PG is still a matter of research. Compton Cameras (CC) have been proposed for PG real-time dose monitoring mainly due to their good time and spatial resolution and their lack of mechanical collimation, which worsen the intrinsic PG high intensity. Conventional CC are based on one or more scatterers and an absorber. The incoming γ direction requires at least two interactions in different layers, assuming the γ energy is known. However, the broad PG energy spectrum in HT makes it impossible to know in advance the energy of the γ. Therefore, three interactions are more suitable to determine the γ Compton cone. Nevertheless, two interactions can be still used, if the incoming energy is included as a part of the reconstruction inverse problem (4D spectral reconstruction). The noise scenario is thus different depending if two or three interactions are employed. For that reason, the expected reconstructed image can be affected differently by spurious data, such as random coincidences and pile-up in the detector from high intensity beams. In this work, a Monte Carlo (MC) based study is carried out with the aim of evaluating the impact of different noise sources (neutro- s, random coincidences and pile-up) for range determination. The complete chain of detection, from the beam particle impinging on a phantom to the event reconstruction, is simulated using FLUKA. The range location is later estimated from the reconstructed image obtained from a 2 and 3 interaction algorithm based on Maximum Likelihood Expectation Maximization (MLEM).