Stereo Video-based Head Pose Tracking for Motion Correction in PET Reconstruction

Typically, movement of the patient during any medical imaging scan will result in a degraded final image if this motion is not accounted for. This research project aims to use the video from two cameras, recording from different angles, to estimate the motion of a patient’s head during a PET (positron emission tomography) scan. By using the video from two cameras to track features of the patient’s head, motion in all three dimensions can be estimated. This motion estimation will then be used to reconstruct an improved final image, by taking into account the patient motion that occurred during the scan. Besides improving the final image, this method holds a benefit over current methods in that it does not require any markers or restraints be placed on the patient during imaging, techniques which are currently used to track or restrict a patient’s motion.

Russell Warneboldt
Faculty Supervisor: 
Dr. Ghassan Hamarneh
British Columbia