TOF-MLAA for attenuation correction in thoracic PET/CT

The TOF-MLAA algorithm (Time Of Flight Maximum Likelihood estimation of Activity and Attenuation) simultaneously estimates the emission image and attenuation coefficient distribution (mu map) in time-of-flight PET performed on PET/CT systems, where the mu map is normally derived from a CT scan. We have implemented a version of this algorithm and evaluated it in human and phantom PET/CT data sets. In two patient studies, TOF-MLAA reduced PET image errors that were due to involuntary motion, and resulted in smoother images in which noise, as measured by region standard deviations, was slightly reduced. We performed a PET/CT study of a thorax phantom with a cardiac insert. The phantom was stationary in one scan and moved during two additional scans. TOF-MLAA responded appropriately when motion occurred. We noted the introduction of diffuse edges in mu maps when motion occurred, which also appeared to be appropriate. In the phantom study, we demonstrated a reduction of PET artifacts that were due to motion-related overcorrection for attenuation in the lower lungs. Because TOF-MLAA is not quantitative, image values in the phantom study were in error by as much as 30 percent, though typically less. We calculated polar map images of the phantom´s cardiac insert. Our experiment failed to introduce a prominent attenuation correction artifact in the heart, and as a consequence TOF-MLAA performed similarly to CT-based attenuation correction, or slightly worse, in the polar map analysis. Polar map errors due to motion were much larger than errors due to attenuation effects. The phantom scan demonstrated that optimal respiratory gating could fix motion related polar-map errors, even when CT-based attenuation correction was used.