Evaluation of the accuracy of PET lung images when uniform attenuation within the lungs is assumed

Objectives Whole body PET MRI systems require the generation of approximate attenuation maps which may be problematic in the thorax. The aim of this study was to investigate the errors associated with using an assumed constant lung attenuation coefficient for correcting PET studies.

Methods Nine clinical PET CT studies were used as mathematical phantoms. Emission projections were generated from the clinical PET studies and attenuation factors were generated and applied to the emission projections from the corresponding CT images ranging from the lung apices to the dome of the liver. In each of the CT slices, the lung attenuation coefficients were sampled and replaced with the slice-specific calculated mean lung attenuation coefficient as might be done for an attenuation map generated from MRI data.. The attenuation factors from these altered images were applied to the attenuated emission projections. Both the true emission projection set with the CT-based attenuation values and the emission set with the approximate attenuation correction were reconstructed with filtered backprojection. The resulting reconstructed images from these two sets were compared.

Results On average, 70 slices through the lung fields for each of the 9 studies were compared. The mean lung attenuation coefficient was 0.024/cm +/- 0.014/cm (58%). There were significant differences in the reconstructed PET lung values when the mean lung attenuation factor was used in the correction with errors ranging from -52% to +69% with similar behavior in every slice.

Conclusions Because of the large variability of lung attenuation coefficients both within and among individuals, using a single lung attenuation coefficient will result in large errors within the lung fields in PET imaging. This could limit the usefulness of PET MRI when the accuracy of PET lung values is important