Improving lesion detection in whole body PET

Whole body positron emission tomography (PET) is a nuclear medicine imaging technique which provides functional images of the metabolism and biodistribution of a radiolabeled tracer molecule throughout a patient's body in a single study. Whole body PET has shown clinical importance in the evaluation of many cancers. In common applications of whole body PET, such as localization of indeterminate lesions or staging of a malignancy, a clinician seeks to detect small, focal lesions during image interpretation.;Several methods for improving lesion detection in whole body PET were investigated in this work. Techniques of accelerating 3-D reconstruction were evaluated in terms of reconstruction time, data storage size, and preservation of reconstructed image resolution. A method for selecting an optimal linear filter function for 2-D filtered back-projection (FBP) reconstruction was investigated. Lastly, a nonlinear, image domain filtering technique was proposed to improve contrast and noise variance in PET images reconstructed by FBP.;Lesion detection performance of human observers was evaluated quantitatively for several methods of reconstructing and processing whole body PET data. Nine expert observers participated in a receiver operating characteristic (ROC) study using phantom images to determine the effects of lesion size, contrast, and signal-to-noise ratio (SNR) on lesion detectability. The increase in contrast and SNR necessary to improve detection performance from random to superior was ascertained. Four nuclear medicine physicians participated in an ROC study comparing 2-D FBP without attenuation correction with 3-D acquisition, segmented attenuation correction, and a Bayesian, iterative reconstruction algorithm. Images evaluated were representative of patients with possible lung cancer. Increased detection performance was noted for all readers with 2-D FBP without attenuation correction compared to 3-D FBP without attenuation correction and 2-D FBP with segmented attenuation correction. Iterative reconstruction improved the detection performance for all readers compared to FBP; however, the measured improvement was small and not statistically significant.;In conclusion, this work serves to quantify lesion detection performance with differing image quality and reconstruction methods. Improved lesion detection in whole body PET should benefit patient diagnosis, treatment, and outcome.