Doppler angle automation may streamline vascular exams
SEATTLE—A technique that automates Doppler angle estimation in vascular ultrasound exams may provide more accurate and standardized outcomes, enhancing the quality of patient diagnosis and treatment outcomes, according to a scientific session presentation at the 2008 Society for Imaging Informatics in Medicine (SIIM) annual conference.
“Modern ultrasound systems offer a graphical interface tool for the operators to specify the Doppler angle for every site of interest for every interrogated vessel during the diagnostic exams,” said Ashraff A. Saad, PhD, who shared the results of research conducted at Washington University in St. Louis on automating this procedure. “This manual process is time-consuming and can be inconsistent among all operators, even within the same radiology department.”
The variation among operators is of concern because the blood flow velocity measurement in ultrasound plays a crucial role in the diagnosis of vessel stenosis and plaque, Saad said.
The researchers’ technique first acquires ultrasound data for preprocessing. According to Saad, in order to deal with the blood flow pulsatility within arteries, the system first captures a number of consecutive raw color Doppler frames that encompass at least one heart cycle. These frames are then temporally averaged to remove the effect of pulsatility, resulting in a single temporal-average image, which is then thresholded using a system’s internal processing for combining the color image with a grayscale B-mode image, he said.
“The resultant binary image is a good representation of the present vessels,” he noted. “However, due to the well-known color bleeding artifact of color Doppler imaging technology, it often suffers from artificially-connected vessels.”
The next step in the advanced visualization image processing technology is a segmentation step to address the problem of disconnecting linked vessels. The team uses a shape decomposition technique designed to decompose the natural objects into meaningful parts that agree with the human visual system portioning, Saad said. A hierarchical partitioning scheme is applied to disconnect the linked objects.
Their technology then applies a skeleton representation method based on a graph-theoretic approach to detect vessel skeletons. The final step of the automation application calculates the Doppler angle for any arbitrary point, which is specified by the ultrasound system operator, within a vessel, Saad reported.
To verify the accuracy of the technology, the researchers compared the automatically detected angles with manual angles settings performed by sonographers with two methods. They used a simulation that presented color Doppler image to the sonographers and had them draw the Doppler angle line manually, according to their perception. In the second method, the sonographers used a prototype of the technology on volunteer human models; first setting a Doppler angle line manually, then applying the automated tool.
“For both types of experiments, the automatic and manual angle results are very similar, exhibiting high correlation coefficients (0.97 for the simulation experiment and 0.99 for the real-time prototype experiment), low means of individual difference (-0.93 degrees for the simulation experiment and 1.66 degrees for the real-time prototype experiment), plus reasonable standard deviations of the individual differences (5.4 degrees for the simulation experiment and 6.44 for the real-time prototype experiment),” Saad reported.
“Modern ultrasound systems offer a graphical interface tool for the operators to specify the Doppler angle for every site of interest for every interrogated vessel during the diagnostic exams,” said Ashraff A. Saad, PhD, who shared the results of research conducted at Washington University in St. Louis on automating this procedure. “This manual process is time-consuming and can be inconsistent among all operators, even within the same radiology department.”
The variation among operators is of concern because the blood flow velocity measurement in ultrasound plays a crucial role in the diagnosis of vessel stenosis and plaque, Saad said.
The researchers’ technique first acquires ultrasound data for preprocessing. According to Saad, in order to deal with the blood flow pulsatility within arteries, the system first captures a number of consecutive raw color Doppler frames that encompass at least one heart cycle. These frames are then temporally averaged to remove the effect of pulsatility, resulting in a single temporal-average image, which is then thresholded using a system’s internal processing for combining the color image with a grayscale B-mode image, he said.
“The resultant binary image is a good representation of the present vessels,” he noted. “However, due to the well-known color bleeding artifact of color Doppler imaging technology, it often suffers from artificially-connected vessels.”
The next step in the advanced visualization image processing technology is a segmentation step to address the problem of disconnecting linked vessels. The team uses a shape decomposition technique designed to decompose the natural objects into meaningful parts that agree with the human visual system portioning, Saad said. A hierarchical partitioning scheme is applied to disconnect the linked objects.
Their technology then applies a skeleton representation method based on a graph-theoretic approach to detect vessel skeletons. The final step of the automation application calculates the Doppler angle for any arbitrary point, which is specified by the ultrasound system operator, within a vessel, Saad reported.
To verify the accuracy of the technology, the researchers compared the automatically detected angles with manual angles settings performed by sonographers with two methods. They used a simulation that presented color Doppler image to the sonographers and had them draw the Doppler angle line manually, according to their perception. In the second method, the sonographers used a prototype of the technology on volunteer human models; first setting a Doppler angle line manually, then applying the automated tool.
“For both types of experiments, the automatic and manual angle results are very similar, exhibiting high correlation coefficients (0.97 for the simulation experiment and 0.99 for the real-time prototype experiment), low means of individual difference (-0.93 degrees for the simulation experiment and 1.66 degrees for the real-time prototype experiment), plus reasonable standard deviations of the individual differences (5.4 degrees for the simulation experiment and 6.44 for the real-time prototype experiment),” Saad reported.