PT - JOURNAL ARTICLE AU - Amirtaha Taebi AU - Gustavo Coelho Alves Costa AU - Emilie Roncali TI - Personalized dosimetry for brain cancer radioembolization: A feasibility study DP - 2021 May 01 TA - Journal of Nuclear Medicine PG - 1582--1582 VI - 62 IP - supplement 1 4099 - http://jnm.snmjournals.org/content/62/supplement_1/1582.short 4100 - http://jnm.snmjournals.org/content/62/supplement_1/1582.full SO - J Nucl Med2021 May 01; 62 AB - 1582Objectives: To evaluate the feasibility of using a custom algorithm (CFDose) to predict the dose distribution during yttrium-90 (Y-90) radioembolization for brain cancer. Glioblastoma multiforme is the most aggressive and fast-growing type of cancer within the central nervous system. In a recent study [1], Y-90 microspheres were administered through the cerebral artery to canine patients with spontaneous intra-axial brain cancers and results indicated the potential of this approach. Transarterial radioembolization, mostly used for liver cancer, is a type of radionuclide therapy where Y-90 microspheres are injected into the arterial bloodstream to irradiate the tumor internally. Accurate pretreatment dosimetry warrants the maximum delivery of the Y-90 microspheres to the tumor while limiting the dose to the surrounding healthy tissues. For liver cancer, we have developed a patient-specific pipeline, called CFDose, which is based on the blood flow simulation inside the vascular network of each patient to estimate the distribution of the Y-90 microspheres and consequently the Y-90 absorbed dose. We have applied CFDose to the optimization of the Y-90 to inject. This new study aims at investigating the feasibility of using CFDose to predict dosimetry for brain cancer radioembolization. Methods: In CFDose, the vascular network is segmented from volumetric medical images such as computed tomography (CT) scans. Blood flow behavior is then studied within the vascular network using computational fluid dynamics (CFD) simulations to predict the Y-90 microsphere distribution inside the vasculature. Eventually, the absorbed dose distribution to biologic tissues is calculated by convolving the predicted microsphere distribution with a Y-90 dose point kernel. In this proof-of-concept study, we selected a head CT angiogram of a patient obtained with a GE Lightspeed VCT machine after administration of 100 cc of intravenous contrast at 4 ml/sec (Fig. 1a). A mask was created to filter the skull from the images. The cerebral artery was then segmented using a fast marching method and extracting isosurfaces at a gray level of 95 HU. Steady-state CFD simulations were carried out in SimVascular. The blood was considered incompressible Newtonian with a density and viscosity of 1.06 g/cm3 and 0.04 g/cm.s, respectively. The vasculature downstream of each outlet was modeled with a 3-element Windkessel circuit. The inlet flow rate, downstream pressure, vascular compliance, and vascular total resistance were assumed to be 250 ml/min, 46.9 mmHg, 0.21 ml/mmHg, and 120 dyn.s/cm5, respectively. The flow field was calculated by solving the Navier-Stokes conservation of mass and momentum equations [2, 3]. Results: Fig. 1b shows the segmented cerebral artery. Inside this computational domain, velocity distribution and vessel wall pressure were calculated from the CFD simulations. Blood streamlines were then determined which can be used to track microspheres (Fig. 2). One challenge to address in future studies is to acquire patient-specific boundary conditions for the CFD simulations since the results highly depend on those. One of the differences between the cerebral artery and hepatic artery which CFDose was originally developed for is the loopy shape of arteries such as the circle of Willis. In future studies, we will investigate other segmentation methods to better extract these loopy vessels. Conclusions: For one patient with no brain cancer, we presented the potential of CFDose in predicting the Y-90 dose distribution for brain cancer. The next step for this feasibility study include i) enroll brain cancer patients currently treated with other methods, ii) segment their cerebral vascular network with a segmentation method more appropriate for the extraction of the cerebral vessels, iii) carry out unsteady CFD simulations with realistic boundary conditions and calculate microsphere distribution in the brain. Research Support: NIH R21 CA237686 (ITCR) and CCSG P30 (NCI P30CA093373).