Key Techniques Research Of Image Reconstruction For Radiation Dose Reduction In X-ray Computed Tomography

X-ray computed tomography(CT)is widely applied in the field of clinical imaging.However,the technique of CT imaging also reveals some considerable defects.The issue attracting a wide concern is about how serious the CT scanning can damage the human body.Because the high energy photons emitted from the X-ray tube can easily break the structure of a molecule,they definitely can be rather harmful for living body.To ensure the quality of reconstructed image,high radiation dose is required to obtain the projection with low level quantum noise.Obviously,to reduce the radiation dose and to promote the reconstruction quality are conflicting.Furthermore,a relatively low contrast is another defect of the technique of CT imaging.For all kinds of soft tissue,such as vessels,fat and organs,the attenuations of them are similar so that to distinguish them will be a real hard task,which gain the difficulty of medical diagnosis.To solve the aforementioned problems,we make a deep study on them,and the main efforts and results are as follow:1.Propose a reconstruction method with high image quality for spectral CT.There are three main innovation point.1)Making use of the correlation among all the channels,the algorithm effectively reduce the radiation dose for spectral CT.2)By introducing a adaptive step length method based on the strong Wolfe condition,the computational efficiency and range of application are substantially promoted;3)For the truncated projection data,if the reference image can offer a global information,the algorithm can realize a interior reconstruction,even the region out of the interest can be perfectly reconstructed.In practice,Spectral CT can effectively promote the contrast resolution.However,it also has to face the problem of high radiation dose.The proposed algorithm can solve the problem by promoting the service efficiency of the radiation dose.During the application,we first need to search for a reference image which is correlative with all channels.Then an object function that is used to calculate the correlation index between the reference image and the target image can be established.An optimal solution of the target image can be found via minimize the minus object function.Furthermore,by getting rid of the DC component,and introducing the patch strategy can further preserve the detailed information of the reconstructed image.During the optimization procedure,we introduce an adaptive method to select the step lengths based on the strong Wolfe condition,which improves the computational efficiency and application value of the proposed algorithm substantially.We use a mass of numerical simulations and real experiments to test the performance of the proposed algorithm.The results demonstrate that it possesses a strong ability of anti-noise,and can preserve the detailed information of reconstructed images.2.Propose the first motion correction algorithm for the interior reconstruction problem with truncated projection.During the correction procedure,the accuracy of the object motion estimation can achieve pixel level by a polynomial constraint.Moreover,this algorithm does not make use of any priori information to guide the motion correction.Based on the redundancy of the projection data,the projection vectors obtained from all the sample angles are considered containing the motion parameters for their corresponding moments.Through densely sampling operations for the motion parameters,the corresponding re-projections in the high dimensional space can be calculated,so that locally linear relations between the re-projections and real projections can be found.Then with the LLE dimensional reduction algorithm,we can calculate the corresponding motion parameters from given projections.To improve the estimation accuracy for truncated projections,we further restrict the LLE results by a polynomial constraint,which forms a constrained LLE based motion correction algorithm.A large amount of experiments demonstrate that the proposed algorithm can exactly estimate the motion parameters of objects and even it is under a noisy circumstance.3.Propose a general image reconstruction algorithm which can efficiently suppress quantum noise of projections.Under the premise of anti-noise,it preserves a plenty of detailed information in the image edge region,and is good at restoring the overall structure of the object.Because low radiation dose generates the projections dominated by quantum noise,a reconstruction method for high image quality according to the type of data noise is proposed.We first give a signal and frequency based noise estimation for the given projections.With the estimated result,the de-noising theory of non-local Bayes can be applied to execute a de-noising operation.After a series of operations on the de-noised projections,such as discrete gradient transform,re-projection,local contrast operation and weighted optimal operation,we get a better set of projection data.Finally,via image reconstruction from the optimal projection data,we get the final output result.A great number of datasets are introduced to test the algorithm performance,demonstrating its feasibility and universality.