Abstract
This work aims to evaluate the impact of lowering PET dose on image quality and diagnostic confidence of PET images reconstructed by filtered back-projection (FBP), standard maximum-likelihood expectation-maximization (MLEM) and advanced MR-assisted methods such as kernel EM (KEM) and maximum a posteriori EM (MAPEM). List-mode FDG PET data from 5 patients were thinned into count levels of 50% down to just 1% of the full data. The impact of post-reconstruction smoothing on clinical rating of standard-dose PET images was also evaluated. An adaptive post-reconstruction smoothing was devised for each reconstruction and dose level. For 50% dose reduction, all methods maintained the same overall image quality as full-dose ones. For dose levels <25%, image quality deteriorated for FBP and MLEM, while KEM and MAPEM permitted this limit to be further reduced down to 5-10% of the standard dose, corresponding to injected activities of 10-20 MBq. Post-reconstruction smoothing levels up to about 8 mm did not notably affect the diagnostic task of assessing broadly varying FDG uptake patterns in the brain, thereby indicating a dose level lower than the standard one could be administered. In conclusion, these preliminary results suggest that current MR-assisted methods may allow notably reduced injected doses for PET-MR, at least down to 10%, for clinical tasks requiring broad pattern recognition, such as used in assessment of dementia.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call