A Preliminary Phantom Study on the Impact of Region of Interest (ROI) Selection on LAP Luna 3D Sgrt Accuracy and Performance
Abstract
Purpose
To quantify the impact of Region of Interest (ROI) selection on the localization accuracy, latency, and stability of a commercial SGRT system. This preliminary phantom study follows AAPM TG-302 guidance to investigate how the geometric characteristics of the input surface data (the ROI) affect the performance of the system under ideal, rigid-body conditions.
Methods
A LUNA 3D SGRT system was evaluated using head and body phantoms to form a contiguous, rigid surface. Following a reference capture, known 5.0 cm shifts were introduced. The system's reported 6DoF corrections were recorded using 15 distinct ROIs across the head, breast, lung/spine, and pelvis. Accuracy was assessed against a ±2 mm / 1° tolerance. System latency and tracking stability were quantitatively measured.
Results
Rotational discrepancy was observed for both a small forehead (2.0°) and small breast (1.3°) ROIs. Translation accuracy deviated for the small breast ROI (1.6mm) and for a large lung/spine ROI shaped like an inverted Y (1.6mm). The top of breast ROI also demonstrated instability with the highest latency (341ms) and multiple real-time data (RTD) interruptions. Conversely, 12 of the 15 ROIs performed within all tolerances. A latency penalty was observed for some large ROIs compared to the baseline ~ 185ms.
Conclusion
The accuracy and stability of an SGRT system are fundamentally dependent on the geometric characteristics of the user-selected ROI and the surface matching algorithm in this case, an Iterative Closest Point (ICP) algorithm. This rigid phantom study demonstrates that suboptimal ROI selection can lead to clinically significant translational and rotational errors, as well as tracking instability. These results establish that rigorous, data-driven validation of ROI selection protocols for each anatomical site is a prerequisite for safe clinical implementation.