Effect of tomographic operator inaccuracies and respiratory motion on PET/CT lung nodule images smearing.

Abstract

In thoracic PET/computed tomography (CT) imaging, uptake foci usually appear smeared because of postreconstruction smoothing and respiratory motion. The aim of the present study was to assess the respective contributions of the reconstruction process and respiratory motion on PET/CT images. Thirty-one pulmonary lesions were studied. Free-breathing PET/CT acquisitions were followed by a 10-min respiratory-gated PET/CT acquisition. Four different reconstructions were performed by combining two different tomographic operators (TOs) (i.e. the geometric clinical system matrix or a system matrix including the detector response) and taking account (or not) of respiratory motion using a previously developed 'CT-based' technique. For each reconstruction method, lesion segmentation was performed with an adaptive threshold. Next, we computed the volume differences between each reconstruction. Finally, we applied a multiple linear model to compute the relative contributions of TO-based and CT-based respiratory compensation to lesion volume. The three groups, combining the reconstruction methods and the respiratory compensation (or not), differed significantly in terms of the volume differences. For all lesions, the full linear model yielded a regression coefficient R of 76.10%. The partial R values were 65.58 and 10.52% for the detector response operator and the CT-based method, respectively. For lesions in the upper/middle lobes, blurring was mainly because of TO (partial R=78.53%), whereas, for lower lobe lesions, smearing was mainly because of respiratory motion (partial R=56.76%). Our results showed that image reconstruction, by TO accuracy, was the main explanatory factor for lesion smearing when considering the chest as a whole. Respiration had a major impact on the lower lobes.