Research On Artifacts Supression Of Micro-CT Based On Epipolar Consistency

Micro-CT imaging is a non-destructive imaging method with high resolution,it can provide internal structure and 3d morphology of the samples with images in micron scale.Thus,MicroCT plays an important role in medical and biological research.However,Micro-CT imaging is susceptible to artifacts,such as geometrical artifacts and respiratory artifacts.Based on the Grangeat formula,the epipolar consistency employs redundancies between cone beam projection data to express the discrepancy between the ideal geometric model and the actual system,which has practical significance for the improvement of Micro-CT imaging quality.In this paper,the research on the removal of geometric artifacts and respiratory artifacts in MicroCT on the basis of epipolar consistency has been carried out.As for the specific contents of research,firstly,the relations between the cone beam data and the 3D Radon transform presented by Grangeat is introduced in this paper;after that,the epipolar consistency condition of cone beam data is demonstrated.On consideration that the epipolar consistency condition is susceptible to noise,the consistency metric with better antinoise performance,less binding and better parallelism has been put forward.Moreover,the sampling method for epipolar planes has been presented in this paper.Based on the consistency metric above,a new respiratory signal extraction algorithm for living animals is proposed in this paper.In this algorithm,the respiratory waveform is obtained by calculating the change of epipolar consistency metric between cone-beam projection data and extracting the change of corresponding respiratory phase.Furthermore,with the combination of this algorithm and the respiratory gating technique,the respiratory artifact can be effectively suppressed.Simulation results show that our algorithm can effectively extract respiratory waveforms under noise conditions,which has good robustness.Results of in-vivo experiments show that,compared with the existing hardware gating scheme,our algorithm can obtain high-quality respiratory waveforms and reconstructed images without the inconvenience of external hardware.Specifically,results of our algorithm has better quality than the hardware gating scheme under the unstable breathing conditions.As the epipolar consistency relies on the accuracy of geometric model,an online calibration algorithm based on it has been presented in this paper.Firstly,the objective function related to the geometric parameters is established by epipolar consistency metric between projection data.Secondly,to solve the objective function,the simplex simulated annealing algorithm and its performance on non-convex functions have been studied in this paper.The numerical simulations indicate that the online calibration algorithm can achieve good accuracy under different imaging conditions and noise environment.The experimental data indicates that the reconstructed images achieved by the algorithm is basically consistent with that of the current off-line calibration algorithm.Finally,GPU parallelization technology has been applied to accelerate the algorithm and improve the operating efficiency.The calibration algorithm proposed in this paper directly utilizes the projection data,and it not only suppresses the geometric artifacts,but also overcomes the dependence on the phantom.Thus,the real-time synchronous online calibration and image reconstruction of the sample can be achieved.