Photons detection in Positron Emission Tomography through Iterative Rule Learning of TSK rule

A correct reconstruction of medical images in Positron Emission Tomography (PET) needs a precise estimation of the position of the incidence photons in the detector surface. The traditional method based on Anger algorithm calculates the position as a polynomial of the intensities. However, it fails to track the true position near the edges of the detector. In this paper Takagi-Sugeno-Kang (TSK) fuzzy rules are used in order to obtain gradual smooth outputs for different situations with a consequent represented in a similar way as Anger does. The algorithm that learns the TSK rules is based on the Iterative Rule Learning approach. The learned knowledge bases have been tested with a set of Monte Carlos simulations of PET photon detection. Results show a good performance of our proposal, which has been compared with other approaches.

keywords: Iterative Rule Learning, Genetic Algorithm, Takagi-Sugeno-Kang, Fuzzy Rules, Positron Emission Tomography;