Three-dimensional movement analysis in digital subtraction angiography: symbol generation from 3-D optical flow

A knowledge-based system to achieve three-dimensional (3-D) optical flow symbolic classification is proposed. Symbols are matched to verbal labels that describe branch motion through an image sequence. A novel numeric to symbolic image transformation is presented, which provides data for the movement interpretation stage. The system is capable of describing and analyzing some dynamic aspects of 3-D artery motion in space over time, such as rotation and translation of some vessels subsets, labeling of expansion and contraction cycles, and classification of homogeneous movements. A test on a 3-D sequence of reconstructed arteries centerlines is presented