Objective To observe the value of improved marching cubes(MC)algorithm based on Monte Carlo sampling and multi-feature analysis for reconstructing head and abdominal CT images.Methods Totally 69 head axial CT images from one patient with glioma and 108 abdominal axial CT images from another patient with liver cancer were retrospectively enrolled,while 98 head axial CT images of males in visible human project dataset were included.Monte Carlo sampling and multi-feature analysis were introduced into MC algorithm.Monte Carlo sampling was used to simulate the uncertainty of voxel values and estimate the probability of voxels belonging to the isosurface,and dynamic thresholds were generated through combining gradient,curvature,local variance and distance term features of voxels.Then head and abdominal CT images were reconstructed with improved MC algorithm,which were compared with those reconstructed with traditional MC and probabilistic MC(PMC)algorithms.Results Compared with traditional MC and PMC algorithms,improved MC algorithm had lower mean square error,higher signal-to-noise ratio,peak signal-to-noise ratio and computation time for reconstructing head and abdominal CT images without significant difference of Dice similarity coefficient nor recall rate,which could keep the structural loss rate and morphological change rate at low level.Conclusion Improved MC algorithm based on Monte Carlo sampling and multi-feature analysis could significantly improve the robustness and edge accuracy for reconstructing head and abdominal CT images,but the computational efficiency and preservation of head structural details still needed to be optimized.

Objective To observe the value of improved marching cubes(MC)algorithm based on Monte Carlo sampling and multi-feature analysis for reconstructing head and abdominal CT images.Methods Totally 69 head axial CT images from one patient with glioma and 108 abdominal axial CT images from another patient with liver cancer were retrospectively enrolled,while 98 head axial CT images of males in visible human project dataset were included.Monte Carlo sampling and multi-feature analysis were introduced into MC algorithm.Monte Carlo sampling was used to simulate the uncertainty of voxel values and estimate the probability of voxels belonging to the isosurface,and dynamic thresholds were generated through combining gradient,curvature,local variance and distance term features of voxels.Then head and abdominal CT images were reconstructed with improved MC algorithm,which were compared with those reconstructed with traditional MC and probabilistic MC(PMC)algorithms.Results Compared with traditional MC and PMC algorithms,improved MC algorithm had lower mean square error,higher signal-to-noise ratio,peak signal-to-noise ratio and computation time for reconstructing head and abdominal CT images without significant difference of Dice similarity coefficient nor recall rate,which could keep the structural loss rate and morphological change rate at low level.Conclusion Improved MC algorithm based on Monte Carlo sampling and multi-feature analysis could significantly improve the robustness and edge accuracy for reconstructing head and abdominal CT images,but the computational efficiency and preservation of head structural details still needed to be optimized.