The goal of this project is to simulate the PET process starting from the emission point, up to the reconstruction point.
Abstract
The goal of this project is to simulate the PET process starting from the emission point, up to the reconstruction point. The simulator serves those who are interested in dealing with reconstruction algorithms. It creates, in a very short time, any wanted number of detected events from a given 3D picture of the chemical concentration. The list of the detected events can then be used for: checking the runtime of the algorithm, debugging and checking it for errors, testing the algorithm for edge cases and etc. All that without the need to build a real PET laboratory or deal with radioactive materials.
The problem (or the background)
The PET (Positron Emission Tomography) process is a relatively new imaging technique; a chemical that emits positrons is injected into a patients’ body and connects to curtain molecules (glucose etc.), the emitted positrons encounter electrons and in the annihilation process a pair of photons is created. A cylindrical scanner, in which the patient is placed, detects the photons and registers the detection events. The list of the events is then fed into a computer in order to run a reconstruction algorithm. The 3D reconstructed picture will, eventually, show the concentration of the chosen molecule in the body, which may be helpful for a medical diagnosis – for instance, high blood concentration in some cases may be identified with the cancer disease.
The solution (or the basic approach)
To implement such a simulator there were a few tasks that had to be dealt with. There was the need in representing the original image defined by the user. Furthermore, as it was previously discussed, the physics of the PET process can be deterministically derived and predicted on the microscopic level. However, the overall picture of the process on the macroscopic level cannot be readily predicted and is therefore of a statistical nature. In order to artificially create data, which is compatible with a real PET process, the fundamental problem that arises is the need to statistically characterize the process.
Tools
We used MATLAB 6.5 on PC computers.
Conclusions
We believe that our PET simulator can actually be useful to people working with reconstruction algorithms, keeping in mind that this is the first implementation ever, of a simulator of that kind.
Acknowledgment
We are grateful to our project supervisor Stas Cherkassky for his help and guidance throughout this work. Ollendorf Minerva Center and the Vision and Image Sciences Laboratory for supporting this project.