Lung cancer is the leading cause of cancer death in the United States Positron emission tomography (PET) imaging has proven to be a sensitive and specific method for detecting and staging this disease. However, respiratory motion causes image blur and inaccurate quantification of the PET images. The aim of this research was to develop and quantitatively evaluate respiratory motion compensation methods for PET imaging.

Experiments were performed using a computer simulated phantom, a physical phantom, and human subjects. Two different respiratory correlated acquisition techniques were evaluated, where the data were “gated” or sorted into data bins associated with different phases of the motion. Three different motion correction methods were investigated. Each of the motion correction methods used the gated data to obtain transformations between each data bin, which could then be used to register each gated image or gated sinogram to a reference phase. These motion compensation methods were quantitatively evaluated by calculating the signal to noise ratio (SNR) and contrast to noise ratio (CNR) for each sphere or lung lesion. The results were evaluated as a function of object size, object contrast, image statistics, and reconstruction algorithm. Various scatter and attenuation correction methods also were applied to the motion compensated PET data for comparison. Finally, respiratory correlated PET/CT images from five lung cancer patients were evaluated.

The motion compensated PET images resulted in improved signal recovery compared to non-respiratory correlated PET images, and often improved most significantly as the number of data bins increased, lesion size decreased, and lesion contrast decreased. Although SNR and CNR were decreased in the gated PET images, there was not a significant loss in SNR and CNR for small, low contrast lesions. Contrarily, SNR and CNR significantly improved in the motion corrected PET images compared to the gated and non-gated PET images The highest lesion SNR and lesion CNR were observed in images reconstructed with ordered subset expectation maximization (OSEM4i8s). The lesion quantification was significantly affected in the motion compensated images when different attenuation maps were applied, and often resulted in spurious signal recovery.