Plan Optimization for IMRT with Flattening-filter-free (FFF) Beams by using Compressed Sensing Technique

Abstract

High dose rate photon beam with flattening-filter-free (FFF) has recently become available in Varian Trilogy MX LINAC for clinical use. The purpose of this work is to develop a novel inverse planning with consideration of the cone-shaped beam fluence for IMRT with FFF beams so that the number of segments and MUs are minimized. FFF beam data measured from a Varian Trilogy MX unit was utilized for dose calculation and optimization. An incident beam was divided into a collection of 0.5x0.5cm beamlets. To accommodate the inherent non-flatness of FFF fluence, a total-variation regularization (TVR) was introduced in the objective function to encourage piecewise constant fluence in the FFF fluence domain, which was mathematically equivalent to working in the conventional flat fluence domain when the beamlet intensities were normalized according to the measured FFF fluence profile. The system was optimized by using MOSEK software package in Matlab. The performance of the proposed method was evaluated by comparing with IMRT plans obtained using conventional beamlet and direct-aperture optimization strategies for two head and neck patients and a prostate patent. Dose volume histograms (DVH), dose distributions number of segments and MUs of each field are compared for the cases. TVR-based inverse planning utilizes the known cone-shaped profiles of the incident beams and provides clinically sensible IMRT solutions with much reduced number of segments and MUs as compared to conventional approaches. Because the TVR CS technique encourages the piecewise constant condition in FFF fluence domain, the fluence map complexity is markedly simplified in the FFF domain. For instance, an inverted cone-shaped isodose distribution can be obtained with only one segment when TVR is used. In this situation, a large number of segments and much higher MUs would be necessary when conventional beamlet-based or direct aperture optimization is employed. For the head and neck cases, the proposed method slightly improves the target dose uniformity and organs-at-risk sparing, but the numbers of segments and MUs are reduced considerably as compared to the beamlet-based optimization without regularization. The level of improvement in the prostate case is found to be modest, primarily due to the fact that the beam is relatively flat for a small target. The final dose distributions are found to be more conformal than that obtained using DAO using 6 segments for each field. This study presents a TVR-based IMRT inverse planning with explicit inclusion of FFF beam profiles. The proposed method provides a clinically acceptable IMRT plans with minimized number of segments or MUs and may find natural application in IMRT with FFF beams.