Fast, Accurate Unsupervised Segmentation of 3D Magnetic Resonance Angiography

摘要: Accurate automatic extraction of a 3D cerebrovascular system from images obtained by time-of-flight (TOF) or phase-contrast (PC) magnetic resonance angiography (MRA) is challenging segmentation problem due to small size objects interest (blood vessels) in each 2D MRA slice and complex surrounding anatomical structures, e.g. fat, bones, gray white brain matter. We show that multimodal nature data, blood vessels can be accurately separated background voxel-wise classification based on precisely identified probability models voxel intensities. To identify the models, an empirical marginal distribution intensities closely approximated with linear combination discrete Gaussians (LCDG) alternate signs using our previous EM-based techniques for precise LCG-approximation adapted deal LCDGs. High accuracy proposed approach experimentally validated 85 real datasets (50 TOF 35 PC) as well synthetic data special geometrical phantoms known shapes.