The overall goals of this project are to: (i) develop quantitative measures of image quality appropriate for whole body fully-3D positron emission tomography (PET) oncology imaging, and (ii) to determine how to maximize these measures by modifying clinical acquisition protocols and procedures for data processing and image reconstruction. The motivation for this work arises from the unique sensitivity of positron emission tomography (PET) for quantitation of tracer uptake associated with abnormal tumor metabolism. The diagnostic utility of PET oncology imaging is often limited in practice by low tracer uptake and low data collection rates, resulting in images with high levels of statistical noise. In previous work under this grant we proposed combining the higher sensitivity of fully-3D imaging with the use of clinically feasible statistical reconstruction methods to reduce noise propagation. This led to the development of the FORE+(AW)OSEM image reconstruction algorithm, which is now implemented on most PET scanners. Current trends in PET scanner design are towards 3D acquisition modes with even shorter patient scan times. This emphasizes the paramount problem of understanding and controlling statistical noise with clinically feasible algorithms in 3D PET imaging. In this competing continuation proposal, our goal is to form an overall model for the chain of image acquisition, processing, and display. That model will be used to show how changing the 3D PET acquisition protocols, data processing, and image reconstruction procedures can improve specific image qualities relevant to clinical tasks. Reducing statistical noise will be addressed both by optimizing the acquisition protocol and modifying the image reconstruction algorithms, within practical clinical constraints. We will test whether both human and numerical volumetric observer studies (with the three standard orthogonal views) reflect clinical task performance more accurately than traditional planar image analyses. This overall approach of understanding the generation, propagation, reduction, and perception of statistical noise is needed to allow objective choices about the tradeoffs inherent in clinical oncology imaging, to realize the full potential of fully 3D-PET whole body imaging and maximize its impact on patient management.
Showing the most recent 10 out of 25 publications
