TH-EF-BRB-06: A Method to Synthesize Hybrid KV/MV Projections for Metal Artifact Corrections in CBCT




Simultaneous utilization of kV and MV projections is a promising approach to correct metal artifacts in CBCT. However, due to inherent differences in attenuation, beam hardening, and scattering properties of kV and MV beams, it is challenging to employ kV and MV projections in the same CBCT set. To address these limitations, a novel method was developed to generate hybrid kV/MV projections, and to correct metal artifacts in CBCT images.


First, a calibration dataset-based method was developed to correct beam hardening and scatter in kV and MV projections, and to convert MV projections to kV-equivalent attenuation maps. In the next stage, hybrid kV/MV projections were synthesized by weighted summation of corrected kV and MV projections; a user-controlled, generalized logistic curve was utilized as the weighting function, and its shape determined the relative weight of kV and MV projections with respect to relative attenuation differences on a pixel-by-pixel basis.


Phantoms with integrated metal hardware were imaged using 125 kVp and 6 MV beams from a linac with on-board KV imaging. CBCT images were reconstructed from KV-only projections and from hybrid kV/MV projections. In kV-only CBCT sets, metal hardware lead to severe artifacts and HU variations. HU values were underestimated by — 400±117 in a Catphan phantom with femoral implants. On the other hand, CBCT reconstructed from hybrid kV/MV projections exhibited significantly less metal artifacts, and the underestimation in HU values was reduced to — 100±73. Similar results were observed in all imaging experiments.


Our method was successful in reducing metal artifacts, and improving CT number accuracy in CBCT images. Consistency of corrections appeared to be robust under diverse imaging conditions. The user-controlled weighting function allowed fine tuning the balance between the level of desired metal artifact suppression and the preservation of soft tissue contrast in CBCT images.