Probabilistic Modeling of Blood Vessels for Segmenting Magnetic Resonance Angiography Images

Abstract

A new adaptive probabilistic model of blood vessels on magnetic resonance angiography (MRA) images is proposed. The model accounts for both laminar (for normal subjects) and turbulent blood flows (in abnormal cases like anemia or stenosis) and results in a fast algorithm for extracting a 3D cerebrovascular system from MRA data. To accurately separate blood vessels from other regions-of-interest, the marginal distribution is precisely approximated with an adaptive linear combination of the derived model and a number of dominant and subordinate discrete Gaussians, rather than with a mixture of only three pre-selected Gaussian and uniform or Rician components. To validate the accuracy of the proposed algorithm, a special 3D geometrical phantom motivated by statistical analysis of the time-of-flight MRA (TOF-MRA) data is designed. Experiments with synthetic and 50 real data sets confirm the high accuracy and reduced computational cost of the proposed approach.