Positron emission tomography (PET) is a radiologic techniques which, when applied in a quantitative mode, is capable of determining the functional metabolic state of tissue in-vivo. As a result, there has been widespread interest over the past decade in the use of PET both as an instrument of medical research and as a diagnostic tool for patient care. Many of the data analysis techniques used in the PET literature are too cumbersome and complex for routine use and, as a result most of the data generated by typical PET study is only interpreted qualitatively. This approach compromises some of the potential strengths of PET for metabolic imaging. Our long-term goal is to improve the quality of information obtained from PET studies. This proposal focuses on the measurement of blood flow and glucose utilization using tracers which have become most widely used in the field ([F- 18]flurodeoxyglucose, [C-11] glucose and [O-15]water). The work is concerned with the development of accurate, efficient and easy-to- use methods for creating quantitative images of tissue metabolism from dynamic PET studies. The approach is based on mixture analysis models in which the time activity curve (TAC) at a given pixel is expressed as a weighted sum of sub-TACs corresponding to homogenous tissues represented at the pixel. With these models, parametric images of tissue are constructed as a weighted sum of parameters associated with the sub-TACs. An early version of this scheme is already beginning to be used in a set of clinically oriented PET cancer studies at our institution.
The specific aims proposed here are: 1) improved the accuracy and precision of the mixture analysis approach to PET quantitation by (i) applying mixture analysis to raw projection data rather than reconstructed images, (ii) integration of multiple planes of data and, (iii) constructing data-dependent methods to optimally select the amount of spatial smoothing; 2) extend these techniques to single-subject, multiple injection PET studies: and 3) incorporate a modeled blood input function to allow improved quantitation of PET data, particularly in situations where blood sampling has been limited. Each of our efforts is specifically motivated by, and will contribute to, on-going PET imaging protocols at our institution. A range of data from realistic computer simulations and physical phantoms, as well as archive human studies, will be used to developed and evaluate the methodology. As the project develops, data analysis tools will be used made available for incorporation into the ongoing patient studies. In addition, the tools will be incorporated into transportable codes and made freely available for use by researchers elsewhere.
| O'Sullivan, Finbarr (2006) Locally constrained mixture representation of dynamic imaging data from PET and MR studies. Biostatistics 7:318-38 |
| O'Sullivan, F; Saha, A (1999) Use of ridge regression for improved estimation of kinetic constants from PET data. IEEE Trans Med Imaging 18:115-25 |
| Maitra, R (1998) An approximate bootstrap technique for variance estimation in parametric images. Med Image Anal 2:379-93 |
| Shaish, A; Daugherty, A; O'Sullivan, F et al. (1995) Beta-carotene inhibits atherosclerosis in hypercholesterolemic rabbits. J Clin Invest 96:2075-82 |
| O'Sullivan, F (1994) Metabolic images from dynamic positron emission tomography studies. Stat Methods Med Res 3:87-101 |
