Research On Optical-CT Dual-Modality Data Fusion Based On Improved Iterative Closest Point Algorithm

Optical imaging detects fluorescent markers by non-invasive methods in vitro,and can reconstruct the distribution of fluorescent substances in the living body according to the optical information acquired by the tissue surface table,thereby displaying biological function information,it has the advantages of high sensitivity,low cost and simple operation.Computed tomography(CT)can provide structural information of organisms.In order to observe biological tissues from both functional and structural aspects,it provides accurate surface fluorescence information distribution for reverse reconstruction of optical tomography,optical and CT dual-modality data fusion is required.The current method is a registration method based on 2D-3D of the marker point,which is inconvenient to operate and is too dependent on the artificial selection of the marker point.Aiming at these problems,this paper proposes an optical-CT dual-modality data fusion method based on the improved Iterative Closet Point(ICP)algorithm,which is applied to Bioluminescent tomography imaging(BLT).The specific research work is as follows:(1)Three-dimensional surface reconstruction of multi-angle two-dimensional white surface data using an improved voxel-wise method.In view of the commonly used voxel-wise multi-angle optical projection surface reconstruction method,the contradiction between voxel subdivision level,reconstruction accuracy and memory occupancy,based on the reconstruction accuracy and efficiency,this paper proposes a three-dimensional surface reconstruction method adapted to the optimal subdivision level threshold is used to reconstruct the surface of the small animal to obtain 3D white light surface data.In order to verify the feasibility of the proposed algorithm for mouse surface reconstruction,a real cylindrical simulation experiment and real mouse experiment were designed.The experimental results show that the proposed algorithm is more accurate than the traditional method in ensuring reconstruction efficiency and can be applied to mouse three-dimensional surface reconstruction.(2)Registration of 3D white light surface and 3D CT surface data based on the improved iterative closest point algorithm.Aiming at the sensitivity of the traditional ICP algorithm to the initial value,the calculation time is long,and the registration accuracy and speed of the different modal and the large residual data are poor,the ICP algorithm based on the bidirectional distance ratio improvement is used for three-dimensional matching.The method first uses the principal component analysis method to perform coarse registration on the point cloud,then uses KD-tree to perform nearest neighbor search to improve the search speed of the corresponding point pairs,establishes bidirectional matching for each point,by calculating the bidirectional distance and the ratio,an exponential function is introduced to calculate the probability value,and based on this,the probability that the point pair belongs to the correct match is judged.Finally,the optimal transformation is calculated by the weighted least squares method,and the final transformation matrix is obtained by the singular value decomposition method.In order to verify the registrationspeed and accuracy of the improved ICP algorithm,the registration experiment of Stanford point cloud data and the registration experiment of white light and CT data were designed.The experimental results show that the registration accuracy and speed of the improved ICP algorithm are significantly better than the traditional ICP algorithm and TrICP algorithm,which can obtain good 3D white light surface and 3D CT surface data registration results,and provide the basis for the final fluorescence mapping.(3)Using the reconstructed 3D white light surface data as an intermediary,multiple fluorescence information fusion and quantization correction are performed to achieve the final optical-CT dual-modality data fusion.Both two-dimensional white light data and fluorescence data are acquired in the same frame,so their projection matrices are the same.The projection matrix is calculated by the white light data,and the fluorescence information is projected and quantized using the obtained projection matrix.In order to verify the effect of the proposed method in data fusion,it is applied to BLT data and micro-CT data fusion.The real mouse experiment is set up.The experimental results show that the proposed method can effectively complete the dual-modality data fusion and facilitate the accurate reconstruction of BLT.