Technique to distinguish signal from statistical noise in PET imaging

The independent subset cross-comparison method (ISCC) subdivides a PET measurement into two or more statistically independent measurements, each of which is iteratively reconstructed in addition to the original data set. A lesion in the PET scan is expected to appear in the same position in each subset image, whereas statistical-noise artifacts are expected to appear in different places. The subset images are presented to a viewer for a comparison that involves the original image and the subset images. The ISCC method was tested in a PET/CT patient scan in which a known small hot spot was computationally inserted. ISCC correctly identified the inserted hot spot as a real object. The ISCC method was tested, using the formalism of localized receiver operating characteristics (LROC) with human observers who viewed 250 image sets from a mathematical simulation. This preliminary LROC analysis was ambiguous, not showing a clear advantage or disadvantage for the ISCC method. ISCC increased the observers´ confidence in positive findings.