Processing of Respiratory Signals from Tracking Systems for Motion Compensated IGRT

In today’s operating theatres, tracking devices are widely employed. One example is the tracking of breathing motion for image guided radiation therapy. For respiratory motion compensation with the robotic CyberKnife® system, the motion of markers on the patient’s chest is measured and related to the tumor motion [2, 3]. However, the measurements are subject to noise that can hamper further signal processing. We established the measurement noise for five different tracking systems. Information on their accuracy can be found in [4]. Based on its distribution we propose to use the à trous wavelet decomposition for noise reduction. To verify the approach, three experiments were conducted. First, artificial noise according to our measurements was added to a sinusoidal signal. The signal was then processed with the noise reduction method and the original and processed signals were compared. Second, an actual sinusoidal motion was simulated using a robot carrying an infrared marker. The actual data measured with an infrared tracking system was processed and compared to the original signal. Third, the wavelet based method was applied to raw chest marker motion data from a patient treated with the CyberKnife®. It is demonstrated that the signal behaves similarly to the simulations and that the proposed noise reduction method is reasonable in practice. Results for least-meansquares (LMS) and Kalman filter prediction on the processed data indicate that the new approach can improve target localization for motion compensated IGRT.