A hybrid 3d reconstruction/registration algorithm for correction of head motion in emission tomography

Even with head restraint small head movements can occur during data acquisition for emission tomography, sufficiently large to result in detectable artifacts in the final reconstruction. Direct measurement of motion can be cumbersome and difficult to implement, whereas previous attempts to correct for motion based on measured projections have been limited to simple translation orthogonal to the projection. A fully 3d algorithm is proposed that estimates the patient orientation at any time based on the projection of motion-corrupted data, with incorporation of the measured motion within subsequent OSEM sub-iterations. Preliminary studies have been performed using a digital version of the Hoffman brain phantom. Movement was simulated by constructing a mixed set of projections in two discrete positions of the phantom. The algorithm determined the phantom orientation that best aligned each constructed projection with its corresponding measured projection. In the case of simulated movement of 24 of 64 projections, all mis-positioned projections were correctly identified. The algorithm resulted in a reduction of mean square difference (MSD) between motion corrected and motion-free reconstructions compared to the MSD between uncorrected and motion-free reconstructions by a factor of 2.7.