A family of linear algorithms for the prediction of respiratory motion in image-guided radiotherapy

Purpose:
Over the last years the CyberKnife system has won a firm spot in clinical treatment of cancerous regions in the body stem without using respiratory coaching, gating or fixation. To achieve this, the excursion of the patient’s chest is recorded and correlated to landmarks obtained during stereoscopic X-ray imaging [3]. This model is 5 subsequently used to guide the robotic arm carrying a linear accelerator to reduce or eliminate respiratory motion. It has been shown that successful prediction of the time series stemming from human respiration to compensate for system latency is indeed possible [1, 2, 4]. We propose a new family of linear prediction algorithms requiring only little computational complexity while yielding highly competitive prediction results.