Intra-frame Motion Compensation in Multi-frame Brain PET Imaging

Inter-frame and intra-frame motion can adversely impact the performanceof dynamic brain PET imaging. Only correcting the former can still result in degradedqualitative and quantitative performance. Meanwhile, patient motion introducesmismatches between transmission and emission data which may lead to incorrectattenuation and scatter compensation in the reconstruction process. As a result,the reconstructed dynamic images may carry erroneous estimates of radioactivitydistribution. We seek a solution to this problem. We investigated the use of iterative deconvolution coupled with a proposeduse of time-weighted averaging of motion-transformed transmission images to correctthe transmission-emission mismatch artifacts in dynamic brain PET images. Weperformed simulations using real-patient motion profile acquired by the infrared PolarisVicra motion tracking device which estimates 3-D motion transformations during PETacquisition. This was followed by frame-based motion correction employing threedifferent transmission-emission alignment strategies: transmission image transformedby (1) mean motion transformation, (2) median motion transformation, and (3) theproposed time-weighted average of motion-transformed transmission images. The results demonstrate that the proposed approach of using time-weightedaveraging of motion transformed transmission images outperforms conventionalmethods by substantially reducing the transmission-emission mismatch artifacts inthe reconstructed images. Coupled with an alignment of the reconstructed framesfor inter-frame motion correction and a subsequent iterative deconvolution approachfor intra-frame motion correction, the resulting motion compensated images showedsuperior quality, considerable reduction in error norm and enhanced noise-biasperformance compared to conventional methods of transmission-emission mismatchcompensation. The performance was consistent across different levels of intra-framemotion, and the algorithm was amenable to different framing schemes. In frame-based motion correction of dynamic PET images, it is feasibleto achieve intra-frame motion compensation using time-weighted averaging ofmotion transformed transmission images coupled with a post-reconstruction iterativedeconvolution procedure to compensate for intra-frame motion.