Semi-automated boundary detection for intravascular ultrasound

Automated boundary detection in intracoronary ultrasound images is hampered by significant time and space variant specular artifacts and incident angle dependent variations in signal intensity from anatomic boundaries. To overcome these problems, the authors have employed a variety of preprocessing steps coupled with minimal operator input to guide an automated boundary detection technique using compensated simulated annealing and linear programming methods. The semi-automated methods work well and provide reproducible estimates of luminal and vessel dimensions under realistic conditions. The methods and preliminary results of such a semi-automated boundary estimation technique are described.<
>