Ultra-low dose CT attenuation correction for PET/CT

Objectives We evaluate methods for ultra-low dose CT-based attenuation correction (CTAC) for PET/CT imaging. The goals are to reduce CT radiation dose in PET/CT scanning and enable respiratory motion compensation methods that require extended duration CT scans.

Methods The ASIM PET simulator and the CATSIM CT simulator were used to replicate clinical PET/CT imaging conditions. CT dose was reduced by reducing mAs, altering kVp and tube filtration. The impact of increased CT noise and artifacts was mitigated by (1) the intrinsically reduced SNR requirements for CTAC, (2) 2D boxcar CT sinogram smoothing (CTSS), (3) application of an adaptive trimmed mean filter (ATMF). A wide range of parameters were simulated with a 20x30 cm elliptical water phantom to evaluate PET and CT image quality metrics (bias, noise and RMSE) as a function of CT dose. Selected combinations were then applied to an NCAT phantom simulation with focal FDG uptake in 6 cm and 1.6 cm soft tissue and bone lesions.

Results For the elliptical water phantom, CT dose can be reduced from >1mSv to approximately 0.2 mSv without increasing PET bias by more than 5%. Increased tube filtration and higher kVp further increased the ratio of PET image quality to CT dose. Narrower spectra and spectra with higher mean energy were generally more dose efficient. The use of 2D CTSS and ATMF allowed additional reduction of CT dose to approximately 0.02 mSv. NCAT phantom simulations showed that by using 11x11 CTSS and ATMF is was possible to keep PET bias within 5% in the hot features while reducing CT radiation dose more than 20 times. However, the optimal amount of CT sinogram smoothing varies with noise level, for example extreme smoothing causes CTAC resolution mismatch and degraded PET quality.

Conclusions Ultra-low dose CTAC, with an order of magnitude reduction in CT dose, is feasible with minor modifications to current PET/CT scanners. This radiation dose reduction can be translated to longer duration CT scans, enabling PET respiratory motion correction methods.

Research Support Supported by NIH grant R01-CA115870