Full 3-D treatment verification using an aggregated kV/MV cone-beam CT technique

Abstract

Full 3-D treatment verification is desirable and yet to be achieved for conformal radiation therapy, especially for IMRT. Conventionally, treatment portals are verified with 2-D portal images which lack of 3-D anatomical information around the treatment volume. The use of MV portal CT for 3-D treatment verification is limited due to excessive radiation. While the gantry-mounted kV CT delivers much less dose for 3-D verification, it lacks the treatment portal information. This dilemma could be resolved by using both kV and MV imaging devices, orthogonally mounted on the linac gantry. The purpose of this study is to develop an aggregated adaptive cone-beam CT technique for full 3-D treatment verification. In this study, we took an aggregated approach to effectively reconstruct CT images with the combination of full kV projections and truncated MV projections. Cone-beam reconstruction was based on adaptive multilevel algebraic technique (MLS-ART). The projections were acquired using a Varian CL21EX accelerator with both a kV imaging device and an aSi500 portal imager mounted orthogonally on the gantry. At present, each MV projection image was acquired using one monitor unit which could be potentially reduced. The low-exposure kV projections were acquired to reconstruct anatomical CT images and the MV projections were added to enhance the anatomical information around the treatment volume in the reconstructed CT images. To minimize the radiation dose, only limited numbers of kV and MV projections were acquired and only the truncated MV projections were acquired to cover the treatment volume. Phantoms for different anatomical sites were used in this study. The effects of total number of projections, the combination of numbers of kV and MV projections, different anatomical sites, and the attenuation coefficient conversion between the MV and kV projections on the quality of reconstructed adaptive 3-D cone-beam CT were also investigated. Preliminary results from phantom studies indicated that adaptive kV/MV cone-beam CT images generated using as low as 9 truncated MV projections (<10 MUs) and 18 full kV projections provide reasonable estimation of 3-D treatment anatomy for treatment verification. The details of the treatment anatomy improve as the projection number increases. Cross-calibration using a CT phantom with multiple density inserts provides proper conversion between kV and MV projections for kV/MV cone-beam reconstruction. Orthogonal dual-image acquisition could also reduce the acquisition time by about 30–50%. The dual-energy adaptive cone-beam reconstruction introduced a new 3-D verification technique with clinically acceptable dosage. It provides not only 3-D anatomy but also the treatment volume information. Based on the 3-D anatomical structures effectively reconstructed using limited number of projections, dose verification and gated cone-beam CT may become possible for real-time 3-D IMRT dose and anatomy verification