Author: James M. Balter, Alexander Moncion, Ikechi S Ozoemelam 👨🔬
Affiliation: University of Michigan 🌍
Purpose: Sequential dual-energy cone beam computed tomography (DE-CBCT) integrated with an online adaptive platform could potentially improve soft tissue visualization for more accurate anatomical delineation during treatment adaptation. While sequential dual-energy imaging traditionally poses challenges due to motion artifacts, modern systems with rapid scanning capabilities present an opportunity to minimize these limitations. This study aims to characterize and compare the image quality of DE-CBCT against single-energy CBCT using the Catphan phantom.
Methods: A material decomposition model was developed, using polymethyl methacrylate (PMMA) and aluminum basis materials, with a calibration phantom providing 484 unique thickness combinations (PMMA: 4-148mm, aluminum: 8-35mm). CBCT projections of a Catphan 604 phantom were acquired using two protocols: low energy (80 kVp, 101.1 mAs, 334 projections) and high energy (140 kVp, 527.7 mAs, 406 projections). Twenty (20) low-energy acquisitions were averaged to match the noise levels in the high-energy images, and sinogram interpolation was used to create matching projections. Virtual monoenergetic images (VMIs) were generated using material-specific pathlengths and reconstructed using the FDK algorithm. Image quality was evaluated using contrast, contrast-to-noise ratio (CNR), and noise metrics.
Results: An enhanced contrast in VMIs compared to single-energy acquisitions was observed. Quantitative evaluation of five different material inserts showed increasing contrast with decreasing energy, while noise exhibited a minimum at approximately 45 keV. This resulted in a maximum contrast-to-noise ratio (CNR) at 45 keV. Relative to the 140 kVp acquisition, VMIs at 35-45 keV showed CNR and contrast enhancements up to 1.8 and 3.0 times respectively.
Conclusion: Sequential DE-CBCT implemented on a rapid scanning system demonstrates the potential to enhance soft tissue visualization through VMI generation. The optimal image quality achieved at 45-55 keV, characterized by improved CNR and contrast, suggests this technique could benefit adaptive radiotherapy workflows, particularly for more efficient and improved target and normal tissue delineation.