An artificial neural net and error backpropagation to reconstruct single photon emission computerized tomography data

At present, algorithms used in nuclear medicine to reconstruct single photo n emission computerized tomography (SPECT) data are usually based on one of two principles: filtered backprojection and iterative methods. In this pap er a different algorithm, applying an artificial neural network (multilayer perception) and error backpropagation as training method are used to recon struct transaxial slices from SPECT data. The algorithm was implemented on an Elscint XPERT workstation (i486, 50 MHz), used as a routine digital imag e processing tool in our departments. Reconstruction time for a 64 X 64 mat rix is approximately 45 s/transaxial slice. The algorithm has been Validate d by a mathematical model and tested on heart and Jaszczak phantoms. Phanto m studies and very first clinical results (In-111 octreotide SPECT, Te-99m MDP bone SPECT) show in comparison with filtered backprojection an enhancem ent in image quality. (C) 1999 American Association of Physicists in Medici ne. [S0094-2405(99)00602-1].