Accuracy and reproducibility of a single-pose image-to-robot registration method for mobile C-arm cone beam CT guided histotripsy.

Purpose: Histotripsy is a focal tumor therapy that utilizes focused ultrasound (US) to mechanically destroy tissue. To overcome visualization limitations of diagnostic US-guidance, C-arm cone beam CT (CBCT)-guided histotripsy is being developed, for which a mobile C-arm could increase accessibility. CBCT-guided histotripsy uses a phantom with a helical fiducial pattern to determine the CBCT-to-histotripsy robot coordinate transformation. This study presents an image-to-robot registration method requiring only one phantom CBCT, evaluated for accuracy and reproducibility using a mobile C-arm.

Methods: The phantom is attached to a robotic arm (replacing the histotripsy transducer) and positioned at isocenter. A CBCT is acquired and image-to-robot registration performed by registering a digital model of the phantom to observed fiducials in CBCT coordinates. Registration was performed by one user (n = 8/day, 2 days) and by 12 different users (n = 4/day, 3 days) with fiducial registration errors (FREs) calculated. After each registration, the transducer was reattached to the robot and a treatment delivered in a multi-layered, agar-based phantom. Directional and target registration errors (TREs) were calculated as directional and Euclidean distances between planned and observed treatments. Directional error inter-day differences for the single-user experiment were evaluated for significance using 2-tailed unpaired Student t-tests. The effect of user variability on variability of FRE and directional error was evaluated for significance using 2-tailed F-tests.

Results: Registrations yielded FRE of 0.12 ± 0.03 mm and TRE of 1.51 ± 0.83 mm. Targeting error significantly increased along the transducer's short axis between days (0.88 ± 0.60 vs 1.43 ± 0.18 mm, p = 0.025) for the single user, with a similar trend for the multi-user experiment (1.45 ± 0.79, 2.70 ± 0.19, and 2.83 ± 0.40 mm). User variability, and thus robot pose variability, did not significantly affect variability of FRE or directional error.

Conclusions: Mobile C-arm CBCT-guided histotripsy showed high accuracy with minimal yet nonnegligible TREs, consistent within but not between days, demonstrating that errors can be measured and accounted for, ideally near treatment day to maximize accuracy.