par Lemaître, Cedric;Bruyndonckx, Peter;Devroede, Olivier;Krieger, I. M.;Tavernier, S. 
;Van Der Laan, J(jan) D.J.;Maas, Marnix M.C.;Schaart, Dennis D.R.
Référence IEEE Nuclear Science Symposium conference record, 5, page (3135-3138), 4179696
Publication Publié, 2006
;Van Der Laan, J(jan) D.J.;Maas, Marnix M.C.;Schaart, Dennis D.R.Référence IEEE Nuclear Science Symposium conference record, 5, page (3135-3138), 4179696
Publication Publié, 2006
Article révisé par les pairs
| Résumé : | PET-detectors based on monolithic scintillator blocks have the potential to significantly improve the sensitivity. The impinging photon position of the 511 keV photons on the detector surface is derived from the scintillation light distribution measured by an APD array. To extract this position information, we used Neural Networks (NN). To this end, each detector module has to be position-calibrated by training neural networks. The neural networks can immediately yield a parallax insensitive incidence position of the photon when it was trained with sample data of photons impinging under a similar angle. Different procedures to obtain the calibration data in an automated way on a fully assembled PET system have been developed and validated on a simulator set-up. © 2006 IEEE. |
