Author: Raneem Atta, Alejandro Bertolet, Mislav BobiΔ, Wesley E. Bolch, Robert Joseph Dawson, Carlos Huesa-Berral, Harald Paganetti, Eric Wehrenberg-Klee π¨βπ¬
Affiliation: Massachusetts General Hospital, University of Florida, Department of Radiation Oncology, Massachusetts General Hospital and Harvard Medical School, Massachusetts General Hospital and Harvard Medical School, Department of Radiology, Massachusetts General Hospital and Harvard Medical School π
Purpose: Representations of intra-organ vasculature have a variety of uses in the field of computational dosimetry but generally rely on models derived from population-averaged reference individuals. This study presents a novel approach to the generation of patient-specific 3D models of intra-organ vasculature, with a specific focus on the liver.
Methods: Six hepatocellular carcinoma (HCC) patients treated with Y-90 radioembolization were selected. These patients received an iodinated contrast agent for cone-beam CT angiography (CBCTA) imaging, enhancing the visibility of major arterial vessels in the liver. Several different image filtering methods were developed and compared to generate accurate vessel segmentations of the CBCTAs, while outer surfaces of the liver were contoured manually in MIM SurePlan Y90 v7.1 by an expert radiologist. Following this segmentation, the liver contour was deformably registered with a polygon mesh model of a reference liver partitioned into the standard eight segments of the Couinaud classification system. Two vasculature network models were then generated, each containing hepatic arterial, hepatic venous, and hepatic portal venous trees: the first was completely synthetic, with vessels generated procedurally within the patientβs liver contour; the second was generated by identifying the endpoints in the existing vasculature segmentation and algorithmically extending the vasculature at the endpoints, perfusing the entire organ.
Results: For all patients, the proposed method was able to construct highly detailed 3D models of intra-liver vasculature which incorporate the major geometric features present in the CBCTA arterial tree segmentation.
Conclusion: The models generated using the described approach are readily amenable to a variety of applications, including integration with stochastic blood particle tracking models, Monte Carlo radiation transport codes, and computational fluid dynamics simulation toolkits.