PUBS: Pulsatility‐based segmentation of lumens conducting non‐steady flow

Abstract

Dynamic velocity‐encoded phase‐contrast MRI (PC‐MRI) techniques are being used increasingly to quantify pulsatile flows for a variety of clinical applications. Studies suggest that the reliability of flow quantitation with PC‐MRI appears to be dominated by the consistency in the delineation of the lumen boundary. An automated method that utilizes both spatial and temporal information has been developed for improved accuracy and reproducibility. The method's accuracy was evaluated using a flow phantom with 8‐ and 5‐mm‐diameter lumens at two different flow rates. The reproducibility of the method was further evaluated with arterial, venous, and cerebrospinal fluid flow data from human subjects. The results were compared with measurements obtained manually by observers of different skill levels. Measurement values obtained manually were consistently smaller than those obtained with the pulsatility‐based segmentation (PUBS) method. For the 8‐mm lumen, significant improvements in measurement accuracy were obtained. Average lumen area measurement errors of about 18% for the high and low flows, obtained manually by a skilled observer, were reduced to 2.9% and 4.8%, respectively. For the 5‐mm lumen, the skilled observer underestimated the lumen area by 13%, while the PUBS method overestimated the lumen area by 28%. Overestimated lumen area measurements for the smaller lumen are attributed to the partial‐volume effect. There was significantly less measurement variability with the PUBS method. An average fourfold reduction in interobserver measurement variability was obtained with the new method. Magn Reson Med 49:934–944, 2003.