Sinogram analysis of aperture optimization by iterative least-squares in volumetric modulated arc therapy

Abstract

Iterative least-squares (ILS) has been used in CT image reconstruction for many years and has also been successfully applied to fluence optimization in intensity-modulated radiation therapy. However, most optimization schemes use aperture optimization, where it is not so clear how to apply ILS. This study therefore proposes a method of using ILS for aperture optimization in volumetric modulated arc therapy (VMAT) and evaluates the performance by comparing with segment weight optimization. The AutoBeam inverse treatment planning system was used to optimize fluence, sequence the fluence into deliverable apertures and then optimize these apertures and their accompanying segment weights. ILS was used for both the fluence and aperture optimization stages. In the aperture optimization, if the dose beyond a multileaf collimator (MLC) leaf was too high, that leaf was closed up slightly, and if the dose underneath a leaf was too low, that leaf was opened up slightly. To evaluate this heuristic, five prostate and pelvic nodes patients were retrospectively planned. To evaluate the performance of the aperture optimization, these plans were compared to those produced using segment weight optimization alone after sequencing. Sinograms showing effective monitor units delivered by MLC leaf pairs over the course of beam delivery were used to elucidate the comparison. It was found that aperture optimization with ILS produced treatment plans which were conformal, with dose deposition easily controlled by a few clinical objectives. Sinogram analysis showed that aperture optimization with ILS produced less leaf motion and less monitor unit variation between control points than when using segment weight optimization only. By converting fluence corrections to aperture shape corrections, ILS can be used as a powerful method for VMAT aperture optimization. Sinogram analysis provides a clear insight into the behaviour of the heuristic.