Direct 4D PET reconstruction of parametric images into a stereotaxic brain atlas for [11C]raclopride

A method which incorporates three key processing stages (kinetic parameter estimation, motion correction and image registration from PET image space to stereotaxic image space) into the maximum likelihood expectation maximization (MLEM) reconstruction algorithm is presented. This approach can be of particular significance in the fields of neuroscience and psychiatry, whereby PET is often used to investigate differences in voxel-wise kinetic parameters (e.g. binding potential (BP) and influx rate constant) between groups of participants which require all images of the kinetic parameters of interest to be registered in a common spatial atlas. In current practice, both kinetic parameter estimation and image registration (in addition to motion-correction) are usually performed post-reconstruction. However, estimation of the kinetic parameters after reconstruction can result in sub-optimal estimates due to inaccurate modeling of the noise. Furthermore, performing motion correction and registration after reconstruction can introduce interpolation effects in the final image and cause image resolution degradation. To include the kinetic parameter estimation and spatial transformation parameters (both for motion correction and registration to stereotaxic space) within the iterative PET reconstruction framework should both reduce the error in kinetic parameter estimates and possibly improve image resolution. The performance of reconstruction was assessed using bias-variance and root mean squared error analyses to quantify differences with conventional indirect reconstruction methods. The proposed method not only delivers better image quality, i.e. sharper images, but also a reduction in bias and in root mean squared error in ROI BP estimates.