Image Reconstruction Using DOI Rebinning and Axial Supersampling for the NeuroEXPLORER

The NeuroEXPLORER (NX) is a next generation SiPM based dedicated brain PET scanner with ultra-high sensitivity and spatial resolution. To provide previously unachievable imaging performance, NX provides both time-of-flight (TOF) and depth-of-interaction (DOI) measurements. DOI measurements add another two degrees of redundancy. The data space for 3D TOF data with DOI is 7-dimensional with four degrees of redundancy. The full list-mode iterative reconstruction including DOI, and data corrections can be too computationally expensive. In this work, we investigate an efficient reconstruction to take full advantage of DOI. We first give two linear DOI equations (DEs) to fully characterize DOI redundancy. By applying the two DEs, we develop a fast DOI rebinning method. For each TOF event with a non-zero DOI value, the two endpoints are mapped to two corresponding endpoints in the cells with DOI bin 0. We define a virtual detector geometry that only contains the thin layer crystals with DOI bin 0. After DOI rebinning, all non-zero DOI events are mapped to virtual detectors. Due to the special U-shaped crystal design for DOI decoding, the effective axial width of the U-shaped crystal is double of the transverse width. DOI measurements can not only reduce parallax errors but also increase effective axial LOR sampling. We introduce axial supersampling by using virtual sub-crystals with half axial width, or even a third, to preserve gained sampling with DOI. For each event with DOI, we apply DOI rebinning, axial supersampling and TOF correction. The rebinned list-mode events are then reconstructed using TOF OSEM. We show from simulated reconstruction that the proposed method produces reconstructed images with improved spatial resolution. In summary, we present an efficient DOI rebinning and axial supersampling to fully incorporate DOI for super-resolution image reconstruction for the NX.