Utilization of raw radio frequency signals of intravascular ultrasound for vascular lumen segmentation and three-dimensional modeling

Intravascular ultrasound (IVUS) is widely used for anatomical assessment of coronary arteries, but its potential for hemodynamic evaluation remains underexplored. As a solution, we propose a method for modeling 3-D vessel geometry using IVUS radio frequency (RF) signals, focusing on flow-specific features to extract the vascular lumen. A 60-MHz single-element IVUS transducer rotating at 180 Hz was used to acquire RF data from a wall-less phantom with a straight, 2-mm radius flow channel. A blood-mimicking fluid was injected at an average flow speed of 10 cm/s, and the transducer was pulled back at 3 mm/s. Flow signals were extracted with singular value decomposition filtering, followed by attenuation correction to enhance lumen boundaries. An active contour approach was then applied for segmentation. Thirty-two consecutive cross-sectional frames were processed to reconstruct a 3-D lumen model, taking into account the pullback speed of the transducer. The proposed method achieved an average IoU of 0.948 for lumen segmentation and produced a 3-D model with a mean radius of 2.04 mm. These results demonstrate the effectiveness of RF-based signal processing for precise IVUS-based vessel modeling that will enable hemodynamic assessments of coronary lesions.