The Comparison of Two Optimization Algorithms for Step-and-shoot IMRT Planning

Abstract

There are two optimization algorithms for step-and-shoot IMRT planning: Beamlet-based inverse planning (BBIP) and Aperture-based inverse planning (ABIP). BBIP works with small number of fields, high number of segments, and constant collimator angle. Segments are subordinated automatically according to MLC position for time optimization (sequencer). ABIP is characterized by pre-defined segments, variable collimator angle for each segment, large number of fields, and low number of segments. The purpose of the study is to compare the optimization algorithms in routine clinical praxis. For comparison of both optimization procedures, we used planning systems Plato Sunrise ITP version 1.2.3 (Nucletron) and PrecisePLAN version 2.11 (Elekta) for BBIP and ABIP optimization, respectively. We made 8 pairs of plans in eight patients with locally advanced head and neck cancer. The plans were designed as the first phase of radiotherapy (PTV1) because of high number of critical structures around the target volume in this region. The treatment plans had to be fully comparable regarding the parameter V90 (target volume included in the 90 per cent isodose) for PTV1, mean dose for both parotids and maximum dose for spinal cord. For plan verification we used classical film dosimetry together with measuring of absolute dose in a defined point within the IMRT phantom. Radiation treatment was performed on a linear accelerator Elekta Synergy with 6 MV photons. The plans created on PrecisePlan were used mainly for higher verification precision. For each treatment plan, we compared number of segments, number of monitoring units and treatment time for a single fraction. Statistical analysis was made using paired t-test with two-tailed p values. Mean number of segments produced by PrecisePlan (ABIP) and Plato (BBIP) was 41 and 98, respectively (p < 0,001). Similarly significant difference was observed comparing the mean number of monitor units - 448 for PrecisePlan, and 1591 for Plato (p < 0,001). The number of MU is in fact the real irradiation time (“beam on”), thus the difference might be important for the operating life of radiation-generating parts of the linear accelerator. The number of MU, however, does not represent the real fraction time. To obtain this, we have to include the time necessary for gantry and collimator rotation as well as lamellae movement. The mean fraction times were 15 min. for PrecisePlan and 19 min. for Plato (p = 0,001). The optimization algorithm ABIP for step-and-shoot IMRT produces fewer segments, fewer monitor units and significantly shortens the total fraction time in comparison to the BBIP algorithm.