Characterizing CNR of super-resolution and sub-resolution PET

In this work, we study the effect of intrinsic detector spatial resolution on the contrast-to-noise ratio (CNR) of reconstructed images. Super-resolution PET uses patient or gantry motion to reconstruct images at a higher resolution than the intrinsic detector resolution. Sub-resolution PET reconstructs images at a lower resolution than the intrinsic detector resolution. Super-resolution can increase the image resolution at lower cost, but at a penalty of CNR loss. Sub-resolution is more expensive, but improves the CNR of reconstructed images. In this study, we characterize the CNR for super- and sub-resolution PET using simulated 2-D non-TOF ML-EM data sets. A 1 mm non-TOF sub-resolution system with 2-D data acquisition improves the CNR of 3 mm lesions by two fold compared to a system with 3 mm instrinsic resolution. Meanwhile, super-resolution has a 4-6 fold CNR penalty.