Non-linear spatio-temporal filtering of X-ray fluoroscopy images

Digital temporal and spatial filtering of fluoroscopic image sequences can be used to restore the quality of images acquired at decreased exposure. Linear filters reduce noise and blur edges. In this study, we examine a non-linear edge preserving, spatio-temporal filter that reduces noise variance with minimal edge blurring. Signal detection and discrimination tasks were performed on static targets using a four alternative forced choice paradigm. For both tasks, an index of detectability, d', was measured for filtered and unfiltered image sequences. Average d' of filtered image sequences was significantly (p<0.05) higher than unfiltered d' for both detection and discrimination tasks. Model fits showed d' improvements of 14% (detection) and 22% (discrimination).