Kalman Filter-Based Head Tracking for Cranial Radiation Therapy with Low-Cost Range-Imaging Cameras

To allow for highly accurate localisation in stereotactic radiation therapy, patients are typically immobilised, equipped with optical markers or observed by stereoscopic X-ray imaging cameras. These methods decrease patient comfort and cause additional exposition to ionizing radiation. Consequently, there is a growing demand for contact-free approaches to compensate for patient motion. In this paper we present a first prototype for tracking head motion during radiation therapy using 3D range cameras. We propose a combination of global and local registration methods. We apply a Kalman movement model to observe head motions. The system's capabilities are analysed using a conventional Kinect-like camera which is mainly known as a low-cost consumer electronics device. Our tests show that the proposed setup has the potential of determining head positions with submillimeter accuracy in real-time. However, the use of low-cost cameras turns out to be problematic as they consistently cause substantial systematic errors in measuring depths.