| As a medium,the light is always used to observe the activity at the level of molecular and cellular organism in the region of fluorescence imaging to conduct qualitative and quantitative study and analysis in vivo.Fluorescent imaging havs a number of advantages,such as non-invasive,low prices,near infrared fluorescence penetration etc.Fluorescence imaging usually locates cancer cells by locating fluorescent markers which are injected into vivo and specifically bind to the target tissue,such as cancer cells.The fluorescence emitted by fluorescent markers will exponentially weakened with the distance of light travel increasing,while the autofluorescence signal will still maintain a constant state,the autofluorescence of biological tissue becomes a big limiting factor.The signal of fluorescent markers will be interfered by the autofluorescence,which may locate the wrong position of the target,even worse may drown the useful signal.To remove the autofluorescence signal,signal preprocessing is needed.Several autofluorescence removal methods based on blind source separation are proposed and the effect of removing the autofluorescence of the simulation data and in-vivo data by these methods are analyzed and studied in this paper.The paper referred to the principal component analysis,sparse non-negative matrix methods.The experimental results show that these methods can remove the autofluorescence and extract the corresponding fluorescence spectrum curve,but the autofluorescence is not completely removed. In order to improve autofluorescence removal performance,this paper studies another method which takes the the regularization into account based on sparse non-negative matrix factorization. A comparative study of the results among these autofluorescence removal methods on simulated data and in-vivo data are compared and analyzed in detail.By comparing and analyzing the results,it is obtained that the regularized non-negative matrix factorization based on sparse constraint works best. |
PDF Full Text Request