Commissioning and clinical outcome assessment of a novel surface-guided radiation therapy (SGRT) system at a C-Arm linear accelerator

Abstract

Surface-guided radiation therapy (SGRT) is now widely used for radiation-dose-free, marker-free patient positioning in modern radiotherapy. We commissioned and clinically implemented a novel SGRT system, LUNA 3D (LAP, Lueneburg, Germany), featuring browser-based operation, GPU-accelerated surface reconstruction, frame rates above 12 Hz, a large field of view, and virtual laser projection. Commissioning included tests of temperature drift, reproducibility, translational and rotational shift accuracy, gantry-related camera occlusion, agreement with cone-beam CT (CBCT), and end-to-end dosimetric performance. Results were evaluated using both an SGRT-acquired reference surface and a CT-derived external surface. Temperature drift remained below 0.4 mm on all axes. With the SGRT reference, maximum deviations were at most 0.3 mm translationally and 0.2 degrees rotationally; with the CT-derived reference, translational deviations increased to 0.8 mm, consistent with systematic bias from the reference surface. Agreement between LUNA 3D and CBCT was within 1.0 mm, and end-to-end testing showed CBCT residuals of 0.9-1.3 mm with 1.2% dosimetric deviation. All results satisfied ESTRO-ACROP guideline criteria. Clinical evaluation of 192 breast and 259 pelvic treatment datasets showed significant improvements after implementation: the breast 3D translational vector decreased by 28.7% from 7.00 +/- 4.35 mm to 4.99 +/- 2.75 mm, and the pelvic 3D rotational vector decreased by 24.0% from 2.31 +/- 0.96 degrees to 1.76 +/- 0.67 degrees (both p < 0.001). These results establish LUNA 3D as a reliable SGRT system that improves routine patient positioning accuracy.

Figures (7)

• Figure 1: (a) Isocenter alignment setup for LUNA 3D system. Geometrical calibration plate with EASY CUBE phantom positioned at the linac isocenter. Both the plate and phantom are aligned with the lateral (cranio-caudal) and longitudinal (left-right) room lasers, while the vertical laser is aligned with the center of the phantom. (b) The daily quality assurance is performed using the geometrical plate alone. LUNA 3D software interface displaying the captured calibration plate position surface and the evaluated accuracy of the cameras' calibrations as well as the isocenter alignment.
• Figure 2: Reproducibility test setup with RUBY phantom. (a) The phantom positioned at isocenter using room laser alignment with surface markings. (b) LUNA 3D software interface screenshot showing the live surface (green) and measured 6DoF values. (c) SGRT-reference: reference surface captured by LUNA 3D. (d) SIM-reference: reference surface derived from the external structure of the phantom's CT scan. The ROI defined for tracking (derivation of 6DoF values) is indicated as yellow surface overlay on the reference surface.
• Figure 3: Translational and rotational shift accuracy testing setup. (a) RUBY phantom positioned on dedicated tilting base providing predefined translational shifts (12 mm vertical, 10 mm longitudinal, 15 mm lateral) and rotational shifts ($1^\circ$ yaw, $-1.5^\circ$ pitch, $-2.5^\circ$ roll) relative to the isocenter position. (b) cropped LUNA 3D software interface displaying the six 6DoF positioning deviations from the reference surface. The deviations between the fixed red virtual lasers indicating the room isocenters and the blue virtual lasers indicating the isocenter defined in the phantom (treatment plan) also reflect these introduced shifts. In clinical practice, these serve as intuitive visual feedback to the user, whose aim is to align the blue lines to the red lines.
• Figure 4: End-to-End dosimetric testing workflow incorporating laser-free positioning with LUNA 3D. Schematic representation of the complete workflow from CT simulation through treatment delivery and dosimetric verification. Steps highlighted in red indicate LUNA 3D-specific procedures. The workflow validates the complete laser-free positioning capability from simulation through treatment.
• Figure 5: Spatial location drift in (a) X-, (b) Y-, and (c) Z-direction measured by each camera pod (Pod 1 to 3) as a function of the corresponding temperature registered by the built-in sensor in the camera pod. All location values are normalized to the first measurement point.