PET Research

Positron Emission Tomography (PET) is an imaging modality based on the detection of photon pairs produced by the annihilation of a positron and electron. By tagging different tracers with positron emitting isotopes, we can then perform a wide range of in vivo functional studies.

Advances in PET instrumentation have greatly improved the utility of nuclear medicine for high resolution imaging of biochemical function. However, the filtered backprojection (FBP) image reconstruction schemes used in most clinical settings are unable to fully realize the potential of this modality as they do not take account of the photon-limited nature of the data or the fact that the coincidence data do not represent perfect collimation between individual detector pairs. As newer PET scanners approach the resolution limits of the modality through the use of smaller detectors and 3D data acquisition, these two factors become increasingly important in limiting the resolution and noise characteristics of the reconstructed images.

In our work we are developing image MAP (maximum a posteriori) reconstruction methods for 2D and 3D PET systems based on a Bayesian formulation. These methods combine accurate modeling of the coincidence detection process with statistical priors on the PET images. The goal of this work is to develop accurate image reconstruction methods that can be used in clinical and research applications of PET and which result in improvements in lesion detection and/or image quantitation in comparison to the currently used techniques. These methods are currently being applied to clinical whole body scanners (the Siemens/CTI ECAT HR+), a small animal scanner (the UCLA microPET system) and prototype designs for dedicated positron emission mammography.

Figure: A pair of slices of a 3D reconstructions from a set of FDG data collected on a CTI EXACT HR+ scanner: at left a MAP reconstruction using our code, at right an equivalent filtered backprojection image of the sae slice.

Figure: Image above shows whole body reconstructions from 2D scans of patient with prostate cancer. Images from left to right are (a) FBP, no attenuation correction, (b) OSEM - no smoothing, (c) OSEM - smoothed with 12mm FWHM Gaussian kernel, (d) MAP - no smoothing, (e) MAP- smoothed with 3mm FWHM Gaussian kernel (all OSEM and MAP images used our accurate PET system model) Data and reconstructions courtesy of Magnus Dahlbom, UCLA.

This work is being conducted in collaboration with Simon Cherry and colleagues in the Crump Institute for Molecular Imaging at UCLA and Magnus Dahlbom and colleagues in the department of Nuclear Medicine at UCLA. This work is supported by the National Cancer Institute under grant No. R01 CA59794).

These pages represent a summary of our work, please see publications for details and bibliographies.