Research On Multi-target Fluorescence Molecular Tomography Reconstruction Algorithm Based On Synchronous Clustering

Fluorescence Molecular Tomography(FMT)is a new technology of optical imaging.It can monitor the lesion from the molecular level to help diagnose the disease.Based on the fluorescence distribution of the animal's body surface,the fluorescence target is calculated by using a suitable reconstruction algorithm.In practical applications,practical problems such as tumor adhesion and cancer cell proliferation require further research on multi-target fluorescence molecular tomography.In the case of multiple targets,localized treatment can be facilitated by providing positioning error and individual distribution information for individual sub-targets.Multi-objective reconstruction and the identification of all sub-targets are very necessary.At present,researchers have designed many reconstruction algorithms.Most reconstruction algorithms have good computational effects in single-target experiments,but no further sub-targets can be determined in multi-target experiments.Different from the traditional reconstruction algorithm,this paper proposes a new multi-targets FMT research ideas.Multi-targets reconstruction is regarded as a clustering problem based on reconstruction results,and the clustering algorithm is introduced into the FMT to improve the reconstruction accuracy and sub-target recognition accuracy.The work of this paper also proves the necessity of multi-target discrimination in FMT.The specific research work is as follows:1)Comparative analysis of multi-objective FMT reconstruction based on classical clustering methods.First,the classical K-means algorithm is introduced into the FMT,and its performance is verified by experiments.The experimental results show that the K-means algorithm is not accurate in the application of FMT.It is too sensitive to the selection of the initial point,and it is prone to false clustering.Then the Self-organizing Maps(SOM)is used to cluster the reconstruction results.The experimental results show that the clustering performance of SOM is better,but the clustering center can't be obtained to further improve the clustering accuracy.The application of classical clustering algorithm in FMT will produce various problems,such as K-means clustering accuracy is not high,SOM can't calculate the class center.2)For the various problems existing in the classical clustering algorithm,we introduce the synchronization clustering algorithm into the FMT as a post-processing algorithm of the reconstruction algorithm and perform cluster analysis on the reconstruction results.For the various problems in the classical clustering algorithm in FMT application,we introduce the synchronization clustering algorithm as the post-processing algorithm of the reconstruction algorithm and perform cluster analysis on the reconstruction results.Each data point is viewed as a phase oscillator,and similar data objects affect,interact and gradually aggregate to form a cluster.Through homogenization and digital mouse experiments,the correctness and stability of multi-targets FMT reconstruction based on synchronization clustering are verified.The experimental results show that the algorithm can improve the reconstruction accuracy of individual targets and calculate the cluster centers without the need to use the number of light sources as prior information.3)Multi-targets FMT reconstruction based on synchronization only uses the three-dimensional spatial information of the reconstruction results,and no fluorescence yield information is used.At the center of the reconstruction result,the node's corresponding fluorescence yield value is high,and the node at the edge corresponds to a lower value.According to such characteristics,the iterative formula of synchronization clustering algorithm is improved.And the spatial information and fluorescence yield information of the reconstruction result are effectively used.Through multiple sets of homogeneous imitation and digital mouse simulation experiments,the clustering ability,stability and correctness of the improved algorithm based on synchronization clustering for multi-targets FMT reconstruction are verified.The experimental results show that the improved algorithm has good stability and accuracy,and can calculate more accurate cluster centers,lower error rate.It is less sensitive to the choice of parameters and more stable.Finally,the application of four clustering algorithms in FMT is compared.The experimental results show that the original multi-targets FMT application based on synchronization clustering and its improved algorithm have better accuracy and stability,and can be correct in different situations.