Coupled level set approach to segment carotid arteries from 3D ultrasound images

In this paper, we described and validated a semi-automated algorithm based on the level set method to segment the media-adventitia boundary (MAB) and lumen-intima boundary (LIB) of the carotid arteries from 3D ultrasound (3DUS) images to support the computation of carotid vessel wall volume (VWV). We incorporated local region-based and edge-based energies for the MAB segmentation, and both local and global region-based energies for the LIB segmentation. The two level set functions are coupled using a boundary separation-based energy to encourage an anatomically-motivated boundary separation between the MAB and LIB. An additional energy term attracts the boundary to pass through anchor points placed by an operator. The algorithm was evaluated with 231 2D transverse images extracted from 21 3DUS images. The algorithm gave a VWV error of 5.2%±3.9%, yielded Dice coefficients of 95.6% ± 1.5%, 92.8% ± 3.2% for the MAB and LIB, respectively, and gave sub-millimeter boundary distance errors. The coefficients of variation of VWV from the semi-automated (5.0%) and manual (3.9%) methods were not significantly different.