Fast optimal beam determination for conformation radiotherapy treatment planning

Conformation radiotherapy, which involves intensity modulation of photon treatment beams, offers considerable advantages compared to conventional radiotherapy, since it has the potential for accurately matching the prescribed and delivered dose distributions, hence enabling the effective treatment of complex tumour scenarios. Associated (inverse) treatment planning methods address a constrained linear optimisation problem involving the determination of intensity modulation functions from the prescribed target dose, optimisation criteria and imposed constraints. Many of the reported inverse planning techniques require a considerable number of iterations for algorithm convergence, making them unattractive for clinical use. This paper reports the significant improvement in convergence time possible using a dynamic relaxation technique applied to a Bayesian optimisation process. Performance of the algorithm is illustrated using a complex concave tumour scenario.