Objective: We evaluate a fully data-driven method for the combined recovery and motion blur correction of small solitary pulmonary nodules (SPNs) in F-18 fluorodeoxyglucose (FDG) positron emission tomography (PET)/computed tomography (CT).
Methods: The SPN was segmented in the low-dose CT using a variable Hounsfield threshold and morphological constraints. The combined effect of limited spatial resolution and motion blur in the SPN's PET image was then modelled by an effective Gaussian point-spread function (psf). Both isotropic and non-isotropic psfs were used. To validate the method, PET/CT measurements of the NEMA/IEC spheres phantom were performed. The method was applied to 50 unselected SPNs <or=30 mm from routine patient care.
Results: Recovery of standardised uptake value (SUV) in the phantom image was significantly improved by combined recovery and motion blur correction compared with recovery-only correction, particularly with the non-isotropic model (residual average error 10%). In the patient images, automated segmentation and fit of the effective psf worked properly in all cases. Volume-equivalent diameter ranged from 4.9 to 27.8 mm. Uncorrected maximum SUV ranged from 0.9 to 13.3. Compared with recovery-only correction, combined correction with the non-isotropic model resulted in a 'relevant' (>or=30%) SUV increase in 47 SPNs (94%).
Conclusions: Correction of both recovery and motion blur is mandatory for accurate SUV quantification of SPNs.