Genetic algorithm as a tool for detection setup optimisation: SiFi-CC case study

Abstract

Objective: Proton therapy is a precision-focused cancer treatment where accurate proton beam range monitoring is critical to ensure effective dose delivery. This can be achieved by prompt gamma detection with a Compton camera like the SiFi-CC. This study aims to show the feasibility of optimising the geometry of SiFi-CC Compton camera for verification of dose distribution via prompt gamma detection using a genetic algorithm (GA). Approach: The SiFi-CC key geometric parameters for optimisation with the GA are the source-to-scatterer and scatterer-to-absorber distances, and the module thicknesses. The optimisation process was conducted with a software framework based on the Geant4 toolkit, which included detailed and realistic modelling of gamma interactions, detector response, and further steps such as event selection and image reconstruction. The performance of each individual configuration was evaluated using a fitness function incorporating factors related to gamma detection efficiency and image resolution. Results: The GA-optimised SiFi-CC configuration demonstrated the capability to detect a 5 mm proton beam range shift with a 2 mm resolution using 5e8 protons. The best-performing geometry, with 16 fibre layers in the scatterer, 36 layers in the absorber, source-to-scatterer distance 150 mm and scatterer-to-absorber distance 120 mm, has an imaging sensitivity of 5.58(1)e-5. Significance: This study demonstrates that the SiFi-CC setup, optimised through a GA, can reliably detect clinically relevant proton beam range shifts, improving real-time range verification accuracy in proton therapy. The presented implementation of a GA is a systematic and feasible way of searching for a SiFi-CC geometry that shows the best performance.