Combined motion blur and partial volume correction for computer aided diagnosis of pulmonary nodules in PET/CT

Objective We present an automated scheme to correct PET max-uptake-values of small to medium-sized pulmonary nodules for motion blur and partial volume averaging. Both effects cause significant underestimation of PET max-uptake-values, particularly in nodules below 25 mm diameter, but nodules up to 75 mm might be affected. This compromises the power of PET for the differential diagnosis of such nodules, in particular benign versus malignant. Thus, correcting PET max-uptake-values has the potential to improve the classification of PET-positive pulmonary nodules. Methods The proposed correction algorithm relies on (i) determination of the actual size and shape of the nodule by segmentation of the nodule in the CT image and (ii) estimation of the effective local point-spread-function in the PET image, taking into account not only the inherently limited spatial resolution of the PET scanner, but also respiratory motion effects. Then the expected under-estimation of the PET max-uptake value in the nodule can be computed by simulation, and the correct PET max-uptake is obtained by multiplication with the correction factor (inverse of the under-estimation/recovery factor). Results Depending on the estimated nodule shape and blur width, the resulting SUV correction factors ranged from 1.0 to 11, with an average correction factor of 3.0, with higher values for smaller nodules. In comparison to SUV correction using a simplified spherical nodule model, the true-shape SUV correction factors were on average 30% higher. The feasibility of the method presented here is indicated by the high correlation between fitted and observed PET image profiles for clinical cases (average 0.995). Conclusion Blur and motion correction factors for standardized PET uptake values may significantly change the differential diagnosis of small pulmonary nodules. Feasibility and stability of the proposed automated combined SUV correction method as well as ease of use of the software tool have been demonstrated by retrospective analysis of real PET/CT patient datasets from clinical routine.