UK audit of the interoperability of ordered-subset expectation-maximisation reconstruction algorithms

Objective In 2002 a UK audit was performed by the Nuclear Medicine Software Quality Group of filtered back projection (FBP) software, designed to evaluate the quantitative characteristics of single-photon emission computed tomography (SPECT). Subsequently, the use of FBP has reduced in common practice, with most guidelines now recommending and using iterative reconstruction. This study aimed to audit ordered-subset expectation-maximisation (OSEM) algorithms in clinical use, acting on the same input data. Methods A computational phantom was devised to evaluate the effect of sphere diameter, position and activity concentration along with an assessment of uniformity and resolution. Additional sections were implemented to evaluate the recovery in photopoenic areas and of small lesions adjacent to active structures. SPECT projections were created from the phantom and placed in the Digital Imaging and Communications in Medicine structures of acquired data from three SPECT camera manufacturers. Resultant projections were reconstructed via six commercial reconstruction platforms and quantitative measures from the above sections compared. Results Across all measures it was found that there was excellent agreement among platforms offering similar reconstruction methods. One platform was found to not offer the ability to perform a true ‘pencil-beam’ OSEM reconstruction and results varied with different manufacturer data supplied. Conclusion While there are differences in how reconstruction platforms process data from different manufacturers, these differences were generally small, with results from the one wide-beam reconstruction method having the largest variation. It would be advisable that users implementing sensitivity-based quantitative SPECT should derive factors for the various combinations of acquisition and reconstruction platforms at their disposal.