The goal of this project is to quantitatively demonstrate improved clinical performance of the latest PET/CT systems using task-specific image quality measures in realistic patient studies These results will be correlated to the performance of simpler metrics that are easier to calculate. Our objective is to either guide a reduction in injected activity or imaging time on these latest systems without sacrificing existing clinical capabilities or to demonstrate new clinical capabilities. The evaluation will be performed as a function of scanner design, image reconstruction, and patient habitus in order to maintain good clinical performance over all patients. Currently, PET imaging protocols are designed based on imaging at or near the peak noise equivalent counts (NEC). While the performance of these scanners has improved, the patient injection protocols have remained constant while the imaging time is reduced. Also, while NEC gives an estimate of the general pixel signal-to-noise ratio (SNR), its relation to clinically relevant metrics is not obvious especially since it does not include the effects of spatial and TOF resolution, and image reconstruction. In this project we perform a clinically realistic study replicating two tasks in oncology: (i) lesion detection and localization common in diagnostic PET, and (ii) monitoring tumor response to therapy with multiple scans. Our embed lesions in clinical patient data prior to image reconstruction, thereby providing a ground-truth for comparison. The proposed work is accomplished through the following specific aims: (1) Develop and evaluate a new simplified NEC metric for optimizing PET images for clinically relevant tasks that includes spatial and TOF resolution effects and test its correlation to a task-specific metric, (2) Perform a comprehensive human observer study for lesion detection and localization, and (3) Perform a ROC study that models tumor response to therapy. In the short term, the results of this work will quantitatively demonstrate improved clinical performance with the latest PET/CT (commercial systems as well as PennPET Explorer scanner), thereby allowing shorter imaging times and/or reduced injected activity. Correlation of the task-specific metric results to simplified imaging metrics will also provide an important tool for future PET system design and image reconstruction evaluation and optimization where complex, task-specific evaluations will not be practical. Clinically, demonstration of good performance with very short scan times will open the possibility of breath-hold imaging with no motion artifacts. Our results will also enable an expansion of PET imaging into new clinical areas where patient radiation exposure is a limiting concern in the use of this molecular imaging modality. We envision serial patient imaging for monitoring treatment response, as well as screening/surveillance studies in high-risk patients, 2 areas of immediate relevance.
In the short term, the results of this work will optimize the imaging protocols for scanners from two commercial vendors and a long axial FOV, state-of-art TOF scanner while long term, the simplified metrics that will be investigated in this study will lead to improved methodology for optimizing new PET system designs as well as image reconstruction algorithms. Clinically, the results of this work will allow an expansion of PET imaging into new clinical areas where patient radiation exposure is a limiting concern in the use of this molecular imaging modality. We envision serial patient imaging for monitoring treatment response, as well as screening/surveillance studies in high-risk patients, to be two such areas of immediate relevance.