In-vivo comparison of real-time 4D ultrasound tracking with electromagnetic transponders in the liver during free breathing

Purpose: The feasibility of ultrasound-guided MLC tracking has recently been shown in phantom experiments but in-vivo validation remains a challenging task since a reliable ground truth is difficult to obtain. Manual annotations in ultrasound show high variability while comparisons with other target localization modalities to date were limited to static setup scenarios. The aim of this study is to enable in-vivo accuracy assessments of ultrasound tracking by simultaneous acquisition of real-time electromagnetic (EM) transponder signals. Methods: 4D ultrasound was acquired during free breathing over 8 minutes post-treatment in an SBRT patient with EM transponders implanted in the liver. The ultrasound probe was fixed with a metal holder, visualizing the transponders inside the patient with a framerate of 12Hz. EM tracking was simultaneously acquired at 8Hz per transponder. Ultrasound tracking was performed offline using fast template matching to localize the transponders. Spatial alignment of the different coordinate systems was obtained by calculating the rigid transform between the 3D tracking trajectories. The 3D root-mean-square error (RMSE) was calculated to evaluate the resulting alignment. Results: Strong interference from the EM system caused varying high-intensity streaks in the ultrasound volumes. An online filter iterating over the previous volumes was developed to facilitate stable tracking. The EM signal acquisition was highly disturbed due to the probe holder position close to the antenna, resulting in only one sparsely tracked transponder. However, six synchronized breathing cycles over 25.7s were obtained. The calculated rigid alignment of the tracking signals was close to expected values from the physical setup and the low RMSE of 0.55mm demonstrates good agreement between ultrasound tracking and the EM reference signal. Conclusion: An approach for in-vivo comparison between real-time ultrasound and EM tracking was developed. Having identified and minimized the mutual disturbance factors, this could provide a valuable tool for future ultrasound tracking validation.