Reconstruction of Compton Scattering Chains with a Likelihood-Based Algorithm : A Monte Carlo Simulation of a Silicon Detector for Medical Imaging Applications

Sammanfattning

A new sensor concept based on a monolithic silicon detector offers a promising Compton imaging alternative to conventional medical imaging techniques such as SPECT and PET. Motivated by the combinatorial problem of ordering Compton scattering interactions, this thesis investigates a likelihood-based reconstruction algorithm. The method is evaluated using Monte Carlo-simulated photon interaction data at 140 keV. The study focuses on reconstruction performance under varying detector conditions, including energy resolution, spatial resolution, and passive detector material. The likelihood-based approach performs well in a medical imaging context and benefits from the known incident photon energy and high detector resolution. The implemented algorithm achieves reconstruction accuracies ranging from 80\% to 99\% depending on chain type and detector resolution. Future work should focus on improving the physical realism of the model, including effects such as Doppler broadening and Rayleigh scattering. Computational efficiency is another important area for further study.