Efficient 3D TOF PET Reconstruction Using View-Grouped Histo-Images: DIRECT - Direct Image Reconstruction for TOF

Abstract

For modern time-of-flight (TOF) positron emission tomography (PET) systems, in which the number of possible lines of response (LOR) and TOF bins is much larger than the number of acquired events, the most appropriate reconstruction approaches are considered to be the list-mode ones, where the processing of the acquired data is done event by event. However, their shortcomings are relatively high computational costs for reconstruction and sensitivity matrix calculation. Furthermore, new sensitivity matrix has to be calculated for each (attenuation) object, and all possible LORs have to be considered for its calculation, not only those which coincide with the actually detected events. Efficient treatment of the TOF data within the proposed DIRECT - Direct Image Reconstruction for TOF - approach is allowed by 1) proper angular (polar and co-polar) grouping of the TOF events to a set of views as given by angular sampling requirements for TOF resolution - each view having a common TOF kernel, and 2) placement/deposit of the grouped events, and all correction data, into the histo-images (one histo-image per view), having the same geometry (voxel grid and orientation) as the reconstructed image. Unlike binning, proper storing operation does not compromise resolution of the data since events are directly stored into the histo-image elements at desired image (voxel) resolution. The reconstruction operations do not negatively affect the resolution neither since no tracing or interpolation operations are needed - all data and operations are directly in the image space. Very efficient reconstruction operations and calculation of correction coefficients can be designed utilizing the image equivalent format and grouping, as demonstrated in this work using DIRECT approach with iterative row-action maximum-likelihood (RAMLA) algorithm.