An image reconstruction method with a locally adaptive gating scheme for PET data.

In conventional gating approaches for positron emission tomography (PET), a single number of gates is predetermined for the whole field of view (FOV) regardless of spatially variant motion blurring effects, which compromises image quality by under-gating regions of large motion and over-gating static regions. To achieve the best resolution-noise trade-off for the whole FOV, we proposed a new approach that incorporates a spatially variant number of gates into gated image reconstruction. The first step was to estimate the motion amplitude of each spatial location. A preliminary set of gated image reconstructions was generated from the PET data. The spatially variant motion amplitudes were approximated based on the registration of 2D maximum intensity projections of the gated reconstructions as well as prior knowledge. Second, the spatially varying motion amplitudes were used to determine the optimal number of gates for each region. Finally, the adaptive gating image reconstruction algorithm that incorporates a gating transform function to model the spatially variant number of gates was applied to generate adaptively gated 4D images. Scans from large FOV systems were simulated using actual multi-bed patient data from a clinical scanner for evaluation purposes. Images reconstructed with the conventional gating scheme as well as static reconstruction were obtained for comparison with the results obtained using our new method. In areas with lower estimated motion amplitudes (such as the spine), the reconstructed images using the new approach showed reduced noise compared to images with conventional gated reconstructions and comparable quality with non-gated images. In areas with large estimated motion amplitudes, such as in the lung and liver, contrast and resolution of images using the new method and conventional gated-reconstructions were comparable, and both were higher than those of non-gated images. The results indicate that using a locally adaptive number of gates based on respiratory motion amplitude instead of a fixed number of gates can improve the statistics of gated PET images by optimizing the local noise-resolution trade-off.