PET-guided adaptive CCO for patient-specific multiscale coronary arterial network generation and hemodynamic simulation

Abstract Background Positron emission tomography myocardial perfusion imaging (PET-MPI) is widely used in clinical medicine to quantify myocardial blood flow (MBF). However, PET-MPI alone does not resolve the dynamic coupling of coronary and microvascular flow and myocardial blood volume; both are central to the pathophysiology of ischemic heart disease. Herein, we propose an adaptive constrained constructive optimization (CCO) algorithm to generate coronary vascular networks from patient - specific whole-heart PET-MPI-derived MBF maps. Methods The proposed PET-derived vascular generation algorithm ingested each patient’s three-dimensional myocardial geometry from static PET, voxel-wise MBF derived from dynamic Rb-82 PET kinetic modeling, epicardial stenosis from fluoroscopy imaging, and a model of the epicardial coronary vessels that served as the model’s starting point. Bifurcating arterial trees were then generated by extending from epicardial vessels toward the endocardium, constrained by PET-derived MBF distribution, physiological branching morphometry, and a rheological model. The method was evaluated in six patients with myocardial ischemia. Results The generated vascular networks produced terminal vessels with approximate diameters of 120 µm. High reproducibility was achieved across random initializations (mean-square-error < 0.02). Vessel diameters showed excellent agreement with published porcine morphometric data (r > 0.99, log–log scale). Simulated tissue-level perfusion closely matched PET-derived values. Relative to remote myocardial tissues, ischemic regions exhibited model reproduced reduction in flow. Mixed-effects analysis revealed moderate to strong correlation between terminal vascular segment count and segmental perfusion (r = 0.60). Conclusion PET-guided, patient-specific coronary microvascular network generation based on our adaptive CCO computational framework enables pulsatile hemodynamic simulation and provides a platform for future in-silico derivation of invasive indices such as coronary flow reserve and the index of microcirculatory resistance across a spectrum of ischemic heart diseases.