The Research Of Cerenkov Luminescence Tomography Algorithm Based On Unsupervised Clustering

Cerenkov Luminescence Tomography(CLT)technology is one of the emerging optical molecular imaging technologies in recent years.CLT technology is based on Cerenkov radiation,but Cherenkov radiation is not a direct product of isotope decay,resulting in lowintensity values and complex transmission processes in biological tissues,which make reconstruction difficult and rebuild locations Deviations from real targets are therefore limited in a wide range of clinical applications.In this paper,we use the unsupervised clustering method to discuss the difficulty of single view reconstruction faced by CLT technology,and the existence of artifacts and noise in the reconstruction results.The specific work is as follows:(1)Aiming at the difficulty of single-view CLT reconstruction and poor reconstruction accuracy,a new CLT reconstruction framework is proposed.The framework is based on a fuzzy C-means clustering algorithm and iterative contraction feasible domain strategy.Each time the reconstructed node energy value is used as the clustering feature value,each reconstruction result is clustered,the light source information is retained,and the useless information is removed.As a feasible area for the next target.In the paper,a variety of simulation experiments and real phantom experiments are carried out to prove the reliability,accuracy,and stability of the method,which can effectively improve the quality of single view reconstruction.(2)Aiming at the problem of artifacts and noise in CLT reconstruction results,a CLT processing method based on artificial neural network is proposed.The method uses automatic encoder neural network,combined with the concept of fuzzy number set and the concept of near four element field,directly processes the reconstruction result.This method only needs to perform one reconstruction calculation,which reduces the time of multiple reconstructions and reduces the error of iteration.Compared with the artificial threshold filtering method widely used in current optical molecular images,the proposed method reduces the influence of human factors of the reconstruction results.In this paper,simulation experiments and various physical phantom simulation experiments in different situations are designed,which proves the versatility of the method of different conditions and provides a new idea about CLT reconstruction.