Estimation of neurotransmitter kinetics from dynamic positron emission tomography data
Abstract
Positron emission tomography (PET) has been used previously to measure the spatial distribution of neuroreceptors and to detect acute neurotransmitter fluctuations in vivo. Conventional data analysis procedures are unable to characterize the temporal pattern of neurotransmitter release. The time course of neurotransmitter concentration may encode pertinent information about brain function and dysfunction, such as learning and memory or drug addiction and abuse liability. We have developed mathematical techniques – collectively called ntPET (neurotransmitter PET) – for the purpose of extracting neurotransmitter kinetics from dynamic PET data. The first of these methods, p-ntPET (parametric ntPET), relies on an enhanced compartmental model and constrained non-linear parameter estimation to simultaneously analyze data from two PET scans (baseline and activation conditions). The p-ntPET model was thoroughly scrutinized by application to realistic simulated data. The results show that p-ntPET is robust to plausible model violations and capable of estimating neurotransmitter release profiles with approximately three minute precision. Analyses of PET data acquired in rats receiving methamphetamine with concurrent microdialysis (direct sampling of extracellular fluid in brain) indicate excellent correspondence between p-ntPET predictions and microdialysis measurements of dopamine release. The second method, lp-ntPET (linear parametric ntPET), is in essence a linearization of the p-ntPET model that utilizes a basis function approach to achieve computational efficiency. Simulation results indicate that lp-ntPET performs similarly to p-ntPET while reducing computational burden by several orders of magnitude. Analyses of baseline-challenge data from rats receiving methamphetamine and single-scan data from a human performing a motor learning task yielded activation profiles that were in good agreement with expected responses. The efficiency of the algorithm will facilitate parametric analysis at each image voxel. Lastly, we demonstrate a method to account for statistical uncertainties in the measured input function. This optimization strategy improves the quantitation of receptor density with standard analysis techniques and may also benefit the performance of the ntPET models. The developments described here enhance the quality of conventional outcome metrics and offer the ability to extract previously unavailable information about the functioning brain.
Degree
Ph.D.
Advisors
Bouman, Purdue University.
Subject Area
Biomedical engineering|Medical imaging
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.