Three Dimensional Segmentation of Intravascular Ultrasound Data

Intravascular ultrasound (IVUS) is a useful imaging technique that can be used to assess the health of coronary arteries. However, manual segmentation of the lumen and adventia boundaries is a time consuming process. Automated methods are needed, but they have to be able to overcome poor signal-to-noise ratios and artifacts commonly present in IVUS images. In this work, three improvements to previous methods were developed and evaluated. These were: (1) a preprocessing step to remove motion artifacts, (2) a new directional gradient velocity term, and (3) a post-processing level-set method. Two IVUS cardiac datasets were used to evaluate the accuracy of the new method over the 3D gradient fast marching method. The new methods, both individually and in combination, were found to significantly lower the volume error.