Stereo computer vision system for measuring movement of patient's head in PET scanning

This paper introduces a stereo computer vision system developed at CPS Innovations for measuring the movement of a patient's head in brain PET scan. A pair of monochrome CCD cameras is installed to acquire the images of the patient's head from two different directions. Simple features (three round dots) on a small target attached to the patient's head are recognized by applying a feature-matching algorithm. The position and orientation of the target is obtained through triangulation calculations. The methods and processes for camera calibration, feature-matching, and measurement accuracy assessment are discussed. The preliminary experiment shows that the root-mean-square measurement error of the system is less than 0.36mm. The system is low cost, compact in size, less intrusive to the patient, and offers robust measurement with sufficient accuracy. It provides an industrial solution for measuring the movement of a patient's head in PET scanning, and can be easily integrated with brain PET scanners such as the HRRT.