ESTRO 2024 - Abstract Book

S4000

Physics - Inter-fraction motion management and offline adaptive radiotherapy

ESTRO 2024

programmed nominal motion or the Identify™ recorded motion. Two metrics were used to test that hypothesis: Pearson correlation (against nominal input motion), Bland-Altman plots in comparison with Identify™.

Results:

The current implementation of proposed system can monitor static and dynamic targets at >10Hz with a measured system latency of 60.73ms (± 0.744ms), qualifying it as a real-time respiratory monitoring system 15 . Only depth based sensing results were available at the time of abstract submission (SI direction/motion). For static targets, the system is stable and accurate at different distances from target (400mm-1200mm): mean absolute error increasing from 0.01mm±0.09mm to 0.79mm±0.96mm for 1min acquisitions. For dynamic targets, the programmed sinusoidal SI motion (25mm amplitude, 20s cycles) was correctly retrieved by both camera system and Identify™ (Fig.1), with the target at approximately 800mm from camera system, and aligned to the Identify™ isocenter.

Since a gold standard for patient respiratory motion cannot be implemented, and phantom experimentation is limited in scope, the Bland-Altman analysis identifies the agreement interval between the two camera systems, allowing neither to be an unequivocally correct system (Fig.2(a)). Hence, for the SI motion presented in Fig.1(a), 95% of the differences of the two camera systems lay within a narrow spread (-1.67mm, 2.48mm), indicating that the proposed system agrees with the clinically utilized Identify™ system. Both systems retrieve motion that strongly correlates with the nominal input motion as seen in Fig.2(b) (p-vals << 0.05).