Fast and automatic heart isolation in 3D CT volumes: optimal shape initialization

摘要: Heart isolation (separating the heart from proximity tissues, e.g., lung, liver, and rib cage) is a prerequisite to clearly visualize coronary arteries in 3D. Such 3D visualization provides an intuitive view physicians diagnose suspicious segments. also necessary radiotherapy planning mask out for treatment of lung or liver tumors. In this paper, we propose efficient robust method computed tomography (CT) volumes. Marginal space learning (MSL) used efficiently estimate position, orientation, scale heart. An optimal mean shape (which optimally represents whole population) then aligned with detected pose, followed by boundary refinement using learning-based detector. Post-processing further exploited exclude cage mask. A large-scale experiment on 589 volumes (including both contrasted non-contrasted scans) 288 patients demonstrates robustness approach. It achieves point-to-mesh error 1.91 mm. Running at speed 1.5 s/volume, it least 10 times faster than previous methods.