PET scanners are used in diagnostic medicine to map the organs of the patient. During the test they inject positron emitting radioactive molecules (called tracer) into the patient’s body and on the resulting image will show where they have been absorbed the chosen molecules. This method is often used to diagnose cancer patients because it is possible to inject such tracer that the cancer cells absorb in a much higher dosage than normal organs do. This way they can not only discover cancer with PET scans but they can also locate it accurately within the patient’s body.
The device is based on the fact that the gamma photons emitted by the injected tracer will have starting velocities in opposing directions while coming from the same point of origin. The photons then get absorbed by the detector-ring around the space of interest and the device registers them as a pair of photons. In ideal cases the point of origin lies on the line defined by the detectors that absorb the photon-pair. As part of the PET scan an intensity map is constructed based on the registered photon-pairs to show the relative concentration of absorbed tracer in the body.
this 3D image is based on the 2D data of the number of photon-pairs registered in a detector-pair. This reconstruction requires the process of huge amount of data while also making the smallest errors possible in the calculations. The algorithms used today do not take the scattering of the photons into account even if the scattered photons can reach the 40% of all the registered photons in case of a human body. In my thesis I participated in the research and development of a method to reconstruct PET/CT measured data while also calculating the effect of photon scattering inside the patient’s body. I mostly worked the implementation and validation of the part that is intended to approximate this effect. The reconstruction is executed by the GPU to increase performance. I will introduce the basics of the reconstruction algorithm used and the physics behind the modeling of the photon scattering and I will show the results of this effort to make PET images more accurate.