Tracing of thin tubular structures in computer tomographic data

Abstract

For many applications in diagnostics and in the planning of surgical interventions, specific structures have to be identified in a patient's volume data set. In this article we give an outline of how the detection of thin tubular structures (e.g., nerves and vessels) can be automated, requiring very little initialization from a human expert. We focused on the nervus alveolaris inferior in the lower jaw and were looking at three details: data acquisition, detection, and validation of accuracy. Our method can be easily adapted to many similar cases such as other nerves, arteries, and veins or bundles thereof.