Positron emission tomography with Compton-scattered photons

Our objective is to demonstrate how the Compton-scattered photons, normally discarded in positron emission tomography (PET), can be successfully utilized for imaging. The scattering in phantoms can be as high as 30-60% of all coincidences. Indeed, there have been few earlier studies to use the scatter data in PET with the energy information of scattered photons. The novelty of the present work lies in developing a methodology using the time-of-flight (ToF) information without the energy requirement. The scattering locus of a photon is known to be circular in a two-dimensional PET. But the annihilation locus is accurately described as elliptical in the present work. The eccentricity of the ellipse increases with increasing pathlength difference of two annihilation photons – one is detected directly while the other is scattered once in the phantom. The technique involves creating many scattering points on its circular locus for a chosen scattering angle, and generating the corresponding elliptical annihilation loci. The procedure is repeated for a range of scattering angles following the Klein-Nishina probability distribution. Depending on the pathlength difference – determined from the known ToF – elliptical loci are created for each event. All annihilation loci are then merged to obtain the activity distribution. We validated our approach using the Monte-Carlo simulated list-mode data sets using single, double, and multiple point sources. The results are very promising as the images are created at the exact source locations with a very good contrast and spatial distribution. However, the spread in the Gaussian intensity profile shows a large width. Currently, our attempt is to reduce the width by using a suitable deconvolution procedure. To conclude, we have established a proof-of-concept imaging procedure utilizing single-scattered photons based on their ToF for the simple activity distributions, but challenges remain for the complex sources.