Coronary Tree Matching Based on Association Graphs with Artificial Nodes

摘要

Tree matching algorithms have various applications in medical imaging for anatomical vessel system such as navigation in the tree structures, planning and treatment procedure or follow-up cardiac therapy. We consider that a dynamic sequence of 3D coronary trees is available from pre-segmented CT data or from reconstruction of the projections acquired from X-ray rotational angiography. We present an automatic graph-based method to find the best match between two non-isomorphic 3D coronary trees that represent two nearly phases of cardiac cycle. Our tree matching algorithm is based on association graphs and maximum clique. In addition to the graph characteristics, the constraint uses also the geometric features of branches (the tree shape). To improve matching performance, our method inserts artificial nodes to match unpaired nodes of the two graphs. We tested our method on real data : 10 volumes of same patient.