Investigation on Parameter Selection of Non-Local Means Filters Using CT Side Information for Multiple I-131 SPECT Scans

Abstract

In our past studies, non-local means (NLM) filtering methods with CT side information for I-131 SPECT estimation have shown promising quantitative results. However, it is challenging to determine optimal filter parameters of these filters without knowing the true image or true dose distribution. Moreover, in SPECT-based dosimetry for I-131 radioimmunotherapy, several parameter optimization steps are necessary since multiple scans are obtained at different count levels to determine time-activity following the tracer and therapy administration. In this paper, we propose practical criteria for choosing filter parameters of NLM filters with CT side information for multiple I-131 SPECT scans without knowing the true activity distribution. We investigate the relationship between quantification and parameters empirically. XCAT simulation results show that our new criteria can determine filtering parameters without knowing the true activity distribution and at different count levels so that the NLM filtering method with CT side information decreased the root mean square error by up to 41% as compared to no filtering while yielding comparable recovery coefficient.