Automated analysis of brachial ultrasound images and image sequences

More than 6 million people in the U.S. suffer from coronary artery disease. Identification of high-risk individuals with subclinical disease is of paramount importance. Growing evidence exists that noninvasive ultrasound imaging of brachial arteries is important in investigating early detection of cardiovascular disease. However, routine assessment of brachial ultrasound image sequences based on manual analysis of long image sequences precludes its clinical utility. We have developed an automated knowledge-based border detection method that combines a priori knowledge about arterial anatomy and ultrasound physics to automatically detect the near and far brachial arterial borders. The method uses graph searching to detect optimal borders in individual frames and ultrasound image sequences. The accuracy of the vessel diameter measurement was validated in 56 individual brachial ultrasound images and in 2 image sequences in comparison to a manually determined independent standard. In the individual frames, the manual and automated vessel diameter measurement correlated very well (R² =0.94, y=0.961+0.20). In the image sequences, the computer-determined diameter changes closely agreed with the expected vessel diameter behavior throughout the cardiac cycle. Our automated vessel diameter measurement method provides accurate and robust diameter measurements and carries a great potential for routine clinical usage and automated analysis of large study populations