3D IVUS border detection in highly diseased arteries with dissecting flaps

Dissecting intimal flaps in highly diseased arteries often create disjointed luminal regions that are 3-D connected to the main lumen in intravascular ultrasound (IVUS) images. These disjointed luminal regions are generally not included as the blood-tissue interface by the expert observer and computer-assisted algorithms. The authors present a 3D connected components algorithm incorporating region-merging that automatically delineates the blood-tissue interface and correctly combines the disjointed luminal regions with the main lumen. The algorithm was quantitatively assessed using linear regression of the cross-sectional areas of the computer-detected and expert-traced contours, and two shape descriptors, the boundary matching descriptor (BMD) and the intersect-to-union ratio (IUR). Results for the slope and intercept were 1.1±0.1 and -2.8±8.0 mm², respectively with R=0.96, n=60. Shape analysis results were BMD=17.3±36.5, IUR=0.85±0.12, n=60