Automated segmentation of coronary wall and plaque from intravascular ultrasound image sequences

The authors report an automated method for segmentation of intravascular ultrasound image (IVUS) sequences that uses a priori knowledge about coronary anatomy and ultrasound physics. External and internal elastic laminae and plaque borders are automatically identified as globally optimal borders using a graph searching approach that detects dual echoes demarcating the coronary wall. The method was applied to six image sequences acquired at a single location in 3 minimally diseased cadaveric left anterior descending coronary arteries subjected to 4 static pressures ranging from 50 to 200 mmHg. A good correlation was obtained between computer identified and observer-defined lumen and plaque areas (r=0.91, y=1.04x-0.50; r=0.83, y=1.08x+0.06; respectively). Border positions were quite accurate with a root-mean-square border positioning error of 0.10/spl plusmn/0.03 mm. The authors´ results demonstrate the feasibility of automated border detection in IVUS coronary sequences.<>