The Study On Diffuse Fluorescence Tomography Based On Adaptive Extended Kalman Filter Technology

Diffuse fluorescence tomography(DFT)combines near-infrared fluorescent agents with diffuse optical tomography to reconstruct three-dimensional distribution of fluorescent targets to reflect physiological information of specific biological tissues,which has the advantages of high specificity,high sensitivity and non-invasive.Dynamic DFT technique developed on the basis of traditional static DFT imaging can dynamically detect rapidly changing fluorescence signals in tissues,and obtain fluorescent pharmacokinetic parameters images combined with pharmacokinetic analysis method of fluorescence agents,which can be used for early diagnosis,grading and treatment evaluation of tumors,etc.Current static DFT reconstruction methods are sensitive to noise,low reconstruction accuracy and low spatial resolution,and it is necessary to select appropriate iteration parameters(such as relaxation factor and regularization parameters)to obtain better reconstruction results.For this reason,we innovatively apply the adaptive extended Kalman filter(AEKF)nonlinear method to DFT image reconstruction in this paper.By introducing the concept of pseudo-dynamic,inverse problem solving is transformed into a dynamic change process,which has the advantages of reducing the sensitivity to noise and no need to choose optimized parameters.In terms of fluorescence pharmacokinetic tomography based on dynamic DFT technique,in view of the advantages of AEKF method,our laboratory has initially developed two methods(indirect method and direct method)to obtain fluorescence pharmacokinetic tomography images based on AEKF.However,how to select the noise covariance matrix in AEKF method,the effect of sampling period on image reconstruction quality,and the ability of indirect and direct methods to reconstruct fluorescent pharmacokinetic parameters when fluorescent agents have different metabolic capabilities are all required further evaluation.Therefore,the main research in this paper includes:First,we developed a static DFT reconstruction algorithm based on AEKF.In order to obtain more accurate reconstructed images,pseudo-dynamics and DFT reconstruction principle were introduced to construct nonlinear state space equations for AEKF analysis.Compared with traditional reconstruction algorithms,numerical simulation and phantom experiments show that this new algorithm can effectively improve the accuracy and spatial resolution of static fluorescence yield images.Healthy mice experiments verify that this new algorithm can still reconstruct the real liver position in living animals and maintain a significant advantage in quantification.Next,we compared the indirect and direct fluorescence pharmacokinetic tomography based on AEKF.First,we select the optimal parameters(noise covariance,sampling period,etc.)suitable for both algorithms by numerical simulation.Based on this,the ability of two algorithms to obtain pharmacokinetic parameters under different metabolic rate was explored,which provided theoretical support for the selection of appropriate analysis methods for different metabolic capacities in the future.Finally,liver metabolism experiments were carried out in mice with liver injury and healthy mice,and the reconstruction results verified the feasibility of fluorescence pharmacokinetic tomography algorithm based on AEKF in living small animals and its ability to evaluate liver function.