Abstract
371
Objectives: Head-motion correction is essential for quantitative high-resolution PET imaging. We quantify motion-correction performance for optical facial-feature-based motion tracking using clinical data from 18F-FDG PET brain studies.
Methods: Twenty 10-minute routine PET/CT brain studies were acquired using a Biograph MCT scanner (Siemens, Germany) using the standard head holder, providing minimal head stabilization. A stereotactic optical camera was fixed to the scanner bore for motion tracking. No markers or fixtures were attached to the patient. PET/CT and optical images (6 frames/sec) were acquired simultaneously with electronic synchronization. Point clouds were created from the optical images and from the CT soft-tissue surface. Point-cloud registration was used with either the CT surface or an early optical image as the reference. The resulting rigid-body displacement and rotation data were used to divide the data into blocks within which the rms motion was less than 0.5 mm (translation) and 0.01 radian (rotation). Motion correction is performed on the sinograms for each block within the reconstruction. Comparison is made of corrected and uncorrected images and the 2D radial autocorrelation and maximum pixel intensities are computed.
Results: The scans were judged technically adequate. However, 65% of the studies yielded more than one movement block, maximum 11. Maximum and rms displacements for the scans were up to 3.4 mm and 1.5 mm respectively. In several studies evident blurring in the uncorrected images is corrected. Reference point clouds based on CT surface data generally are found to provide better registration than optical point clouds. The autocorrelation shows a steeper and narrower peak for the motion-corrected images demonstrating decreased blurring. The motion-corrected images demonstrate larger maximum voxel intensities compared to non-motion-corrected images (paired t, fractional increase 0.22, p-value: 0.029).
Conclusions: Most routinely performed PET brain scans contain enough motion to be usefully corrected by a computationally efficient optical facial-feature-based motion tracking system. Motion correction results in decreased blurring as reflected by visual and quantitative measures. The utility of motion correction is increasingly important as PET image resolution improves and quantitative evaluation becomes a standard part of practice.