Using density computed tomography images for photon dose calculations in radiation oncology: A patient study

BACKGROUND AND PURPOSE: Conventional workflows for dose calculations require conversions between Hounsfield Units (HU) and the mass or electron density for Computed Tomography (CT) images in the Treatment Planning System (TPS). These conversions are scanner– and mostly kVp–dependent. A density representation or reconstruction at the CT level can potentially simplify the workflow. This study aimed to investigate the agreement between these two methods for patients and different calculation algorithms. MATERIALS AND METHODS: Density conversions for conventional HU–density conversions were first established using two phantoms with appropriate inserts. Next, the differences in density and dose calculations between both methods were assessed using 95% Limits of Agreement (LOA) Bland–Altman analysis for 44 consecutive clinical patient cases. These cases represented a mix of indications, algorithms (collapsed cone, convolution superposition, ray tracing, finite–size pencil beam, and Monte Carlo), and scan kVp (80 to 140) in two different commercial TPS. RESULTS: No statistically significant bias in density or dose calculations was found between the two methods. Furthermore, 95% LOAs between both methods were ±0.05 g/cm(3) and ±0.1 Gy for density and dose, respectively. Small but clinically irrelevant dose differences were found in high–density gradient regions for convolution superposition calculations or CT scans with non-delayed contrast agent injections with targets nearby vessels. CONCLUSIONS: The in vivo density–reconstructed images at the CT level were assessed to be equivalent. Therefore, they can simplify and improve clinical workflows, allowing patient–specific acquisitions for contouring and density–reconstructed images for dose calculations.