Author: Yasaman Anbari, Srinivas Cheenu Kappadath, Benjamin P. Lopez 👨🔬
Affiliation: University of Houston, UT MD Anderson Cancer Center 🌍
Purpose: To improve lung dosimetry for 90Y-radioembolization planning, CT-based techniques have been proposed for patient-specific lung mass estimation instead of assuming nominal 1 kg lung masses. Here, we evaluate the robustness of CT-based lung mass estimation using routine chest CT acquisitions and reconstructions.
Methods: All clinical chest CTs available (including breath-hold and free-breathing scans) for 11 patients that underwent 90Y-radioembolization were automatically contoured using Contour ProtégéAI+ (MIM Software 7.3.4). Left and right lung mass densities were individually calculated using the lung mean-HU (LUNG) and soft-tissue mean-HU (SOFT): (LUNG+1000)/(SOFT+1000)*1.03 g/mL. Lung masses were calculated by multiplying the individual mass densities by the respective contour volumes. In studies with multiple reconstructed images, contouring was performed separately in each reconstruction, and lung masses were estimated from every combination of input image series and output contour objects (i.e., single CT scan with 3 reconstructions = 9 mass estimates).
Results: Ninety-five CT studies from multiple institutions and manufacturers and 258 reconstructions were analyzed. Average (median) patient-specific lung masses were 826-1154 (984) g, with lung densities 0.15-0.39 (0.21) g/mL. Across all available acquisitions per patient, the median coefficient of variation (COV) in lung mass was 8.1% (range 3.3-12.1%), reflecting the variability from acquisition parameters. Within a single study, the median COV in lung mass was 4.7% (range 1.0-15.9%), reflecting the variability from reconstruction parameters.
Conclusion: While CT-based approaches can provide patient-specific lung mass estimates, its precision was only 10-15% depending on acquisition and reconstruction parameters used. Use of a nominal lung density (e.g., 0.3 g/mL) with patient-specific lung volumes is erroneous. Standardization of lung mass estimation is needed to understand the propagation of mass uncertainties to lung dose estimates. Analysis to quantify the impact of acquisition and reconstruction parameters (contrast, kVp, slice thickness, voxel size, filter) on lung mass in larger patient cohort is underway.