EMrecon is an Expectation Maximization Based Image Reconstruction Framework for Emission Tomography Data. The EMrecon project was started in 2006 as an open reconstruction software for the quadHIDAC wirechamber small animal PET scanner. It was lead by Thomas Kösters with Klaus Schäfers and Frank Wübbeling. After Thomas left for NYU/Siemens Healthineers, project leadership in Münster was taken over by Dirk Mannweiler. It was funded by the SFB MoBil, the European Institute for Molecular Imaging, and the applied math institute of the University of Münster.
The program was built upon a fast OSEM parallel reconstruction algorithm and included corrections for physical effects, e.g. attenuation, scatter, randoms and motion, as well as support for resolution recovery and list mode. The framework allowed us to test in practice extended reconstruction algorithms, like TV (Total Variation) and other nonlinear regularization schemes. The code was then extended to a broad range of different scanner geometries, among these are Siemens Biograph Sensation 16, mCT, mMR and Inveon small animal PET, and the MEDISO nanoScan in list mode. External scanner definitions, e.g. imported from the GATE simulation toolbox, can be used to adapt the software to general crystal based designs.
The basic reconstruction algorithm is EM (Expectation Maximization) or OSEM (Ordered Subset Expectation Maximization). Variational penalty terms can be included. Extensions allow to estimate motion (via gating, data driven or with external signals) and to correct for this motion in the reconstruction process (model-based approach). For all modalities, listmode and sinogram/michelogram input is supported. EMrecon is implemented for parallelization on Shared Memory-Systems (multicore). Newer developments include TOF (time of flight), angular meshing and fully quantitative reconstruction for the mCT.
Source Code: Email (upon request)
Tutorial: Email (upon request)