Refining imaging parameters for dual-energy cone-beam computed tomography in image-guided radiation therapy.

PURPOSE: To develop a dose estimation method for virtual monoenergetic images (VMIs) derived from sequential dual-energy (DE) cone-beam computed tomography (CBCT), enhancing image quality while reducing imaging dose. The goal is to generate VMIs with lower imaging dose than standard CBCT for future clinical applications.

MATERIALS AND

METHODS

Normalized air kerma (Kair) was measured using an ion chamber for eight CBCT datasets with varying exposures and framerates at 80 and 140 kVp. Correlations between Kair and measured cone-beam dose indices (CBDI) were established to estimate Kair-based imaging dose for DE-CBCT, and then the estimates were subsequently validated. Separately, VMIs were reconstructed from eight new DE-CBCT protocols of a Catphan 604 phantom using the Feldkamp-Davis-Kress (FDK) algorithm within the open-source TIGRE. Image quality of these VMIs at 60 keV, optimal for soft tissue contrast, was evaluated using the contrast-noise-ratio (rCNR) relative to the clinical Pelvis Large Protocol (PLP), the mean Hounsfield units (HU) accuracy over all material inserts, and HU uniformity.

RESULTS

The average difference between estimated and measured Kair was 0.7 ± 2.1% for 80 kVp and 1.3 ± 1.9% for 140 kVp. All VMIs exhibited rCNR values greater than 1 (range: 1.11-1.70), indicating enhanced soft tissue contrast compared to the PLP. The estimated relative Kair for these VMIs ranged from 60% to 100% of a single PLP. VMIs also exhibited improved HU accuracy, reduced HU variance, and substantially improved HU uniformity including those with 60% of the PLP imaging dose.

CONCLUSION

This pilot study demonstrates that VMIs can improve CNR and HU uniformity while reducing imaging dose by up to 40%, relative to the PLP, without compromising HU accuracy. Our approach offers potential for optimizing VMIs by balancing image quality enhancement and dose reduction. Future work will focus on the application of advanced reconstruction algorithms to further improve VMIs quality.