Room: Track 3
Purpose: In contrast to other image guidance techniques, ultrasound offers the distinct advantage of providing volumetric image data in real-time (4D). Matrix array probes have been shown to be faster and potentially better suited for real-time tracking tasks. The goal of this study is to perform the first quantitative comparison of different 4D ultrasound systems with real-time streaming capabilities to assess their tracking performances.
Methods: Three 4D ultrasound systems with matrix probes were investigated for comparable volume sizes at 16cm depth: (1) GE Vivid7, 3V-D probe, volume size 60°x25°, (2) GE Vivid E95, 4Vc-D, 60°x40°, (3) Philips Epiq7, X6-1, 60°x45°. Volumetric data was streamed via Ethernet to an external computer. A robotic arm (Franka Panda) moved the probes relative to an ultrasound phantom with sinusoidal motion patterns of different amplitudes (10mm/35mm) and period lengths (2.5s/5.0s), representing typical liver motion ranges. Robot positions and ultrasound data were acquired for 1min and stored for retrospective analysis. The position of a spherical target was detected in the ultrasound volumes with template matching and compared to the corresponding robot positions by calculating the mean root-mean-squared error (RMSE), with and without compensating for latency between both systems.
Results: All systems continuously followed target motion with framerates of 16.7Hz (Vivid7), 14.6-15.7Hz (E95) and 3.0-9.0Hz (Epiq7). With latency compensated, Vivid7 achieved the lowest RMSE of 0.6±0.1mm, followed by 0.7±0.3mm (Epiq7) and 0.9±0.5mm (E95). Considering a mean latency of ~200ms, RMSE values increased to 2.4±1.6mm (Epiq7), 2.7±1.5mm (E95) and 2.8±1.6mm (Vivid7). Faster target motion, larger amplitudes and higher overall latency (slower framerates) led to an increase in RMSE up to 395%.
Conclusion: 4D ultrasound systems with matrix array probes can track fast breathing motion with sub-millimeter accuracy. Higher system latency and motion amplitude substantially increased tracking errors and should be compensated with prediction algorithms for respiratory motion compensation.
Funding Support, Disclosures, and Conflict of Interest: The 4Vc-D probe was provided by GE healthcare.