A fuzzy artificial neural network-based method for Cerenkov luminescence tomography

Cerenkov Luminescence Tomography (CLT) is a non-invasive three-dimensional in vivo detection technology. However, due to the ill-posedness of CLT, the reconstructed result has many artifacts, which will mislead the researchers to make a wrong diagnostic decision. Enlightened by the development of artificial neural networks, we proposed a Fuzzy Autoencoder Clustering method to eliminate these artifacts and improve the reconstruction quality. To assess the performance of our method, several numerical simulation experiments and real physical phantom experiments are conducted. Compared with the raw reconstruction results and the commonly used manual threshold processed ones, it is demonstrated that our method is capable of filtering the artifact areas effectively, making reconstruction results clearer. It is anticipated that the method presented in this paper will help advance the CLT technology and promote the clinic translation of CLT technology.