Phantom Evaluation of a Time Warping Based Automated Arterial Wall Recognition and Tracking Method

Abstract

Ultrasound-based arterial wall recognition and tracking methods in the literature apply to two-dimensional ultrasound data, either in the form of B-mode images or M-lines radio-frequency (RF) data. We propose a robust dynamic time warping method that is applicable to just one-dimensional single scan-line RF signals. It uniquely analyses the time-varying effects of tissue dynamics on the amplitude and phase features of the RF signals. Its performance was investigated via systemic in-vitro experiments on a pulsatile flow phantom. The recording was performed by an ultrasound imaging system where the B-mode video clips and the raw RF data were saved simultaneously for direct comparison of the proposed method’s versus B-mode reference measurements. The noise of different levels was added to the RF signals to evaluate the method’s robustness. The method detected the arterial walls in 95% -100% of the frames (with SNRs ≥ 10 dB), and for ~100% of those detections, the method accurately localized the walls in the frames. Even when SNR levels were poor (0 dB < SNR < 5 dB) the detection and correct rates were greater than 80% and 90%. The performance figures were consistent for different pulsation rates (0.4 to 3 Hz) emulated. Further, the tracking errors were < 5% for frames with SNR ≥ 5 dB, which improved (errors < 3%) with an increase in SNR. The distension measurements resulting from tracking were repeatable over continuous pulsation cycles (CoV < 0.5%) and were accurate compared to B-mode measurement, with RMSE = 22 μm. The measured versus reference distensions strongly correlated (r = 0.99, p < 0.05) to each other and yielding insignificant (p = 0.17) difference of -6 μm. The method has the potential to facilitate an automated framework for A-mode-based structural and functional analysis of the blood vessels. Therefore, it allows the realization of advanced and cost-effective real-time A-mode systems.