The objective of this project is to develop methodology for energy-based scatter estimation that can be applied to clinical data and produce accurate quantitative PET images over challenging imaging situations such as low collected counts and/or data acquisition at high count-rates. The goal is to enhance the accuracy of PET imaging in situations where current state-of-art scatter estimation techniques are limited in accuracy or perform poorly. In this proposal, we develop scatter estimation methodology that makes full use of the annihilation photon energy information present in the emission PET data together with a simple energy calibration acquired with a physical phantom. We implement, optimize, and evaluate this algorithm on measured data from a clinical PET scanner over all imaging protocols. Our final goal is to implement it on clinical PET/CT and evaluate its impact on real patient data. The proposed work will be accomplished through the following specific aims: (i) Using realistic Monte Carlo simulations to fully implement the proposed algorithm as it will be applied to measured data, followed by parameter optimization and evaluation in reconstructed images, and (ii) evaluating the methodology on measured data (phantoms as well as patient studies) followed by its extension to high count-rate data acquisition situations. In addition to its advantages over existing scatter estimation methodology in situations with low collected counts and/or data acquisition at high count-rates, the proposed technique is expected to be faster and also does not require a transmission or CT image. Successful demonstration of this technique may significantly expand the application of quantitative PET/CT in oncology areas such as treatment monitoring with low-dose repeat PET scans, imaging with new biomarkers that use low positron yield radionuclides (e.g. 124I, 86Y, etc.), or acquiring data at high count-rates (as in cardiac imaging or imaging with 124I or 86Y). Beyond oncology, it will also provide improved quantitation in cardiac studies (82Rb, 13NH3, or 11C-actetate). Since, the proposed scatter estimation method does not require a CT image it may have an application in PET/MR imaging as well as clinical studies with some patient motion ? both situations where the CT image is either not available or is compromised leading to errors in the traditional way of estimating scatter.
Quantitative PET biomarker imaging promises to play a significant role in delivering precision medicine for cancer patients, since it can provide accurate biomarker uptake measurements that are necessary for tumor characterization as well as measuring changes in response to therapy. While PET images acquired during routine 18F-FDG imaging provide highly accurate images, it is challenging to maintain the same performance in situations where few coincidence events are collected (e.g. imaging at low injected activity levels for repeat scans or imaging tracers with a low positron yield) or if the data acquisition count-rate is high (e.g. cardiac imaging or imaging tracers such as 124I or 86Y which produce high background gamma events relative to the positrons). For PET to fulfill its role in this era, quantification accuracy of PET images needs to be maintained over a large range of data acquisition protocols, where the quantification challenge is the accuracy of the scatter estimation method that is used to compensate (correct) for the bias present the data due to scattered coincidences.
