Multi-mode Medical Image Registration System Based On Mutual Information Design And Implementation

With the continuous development of medical imaging,medical image registration has gradually become an indispensable important part in the field of medical image processing.Because the information reflected by various medical images is not the same,the medical images obtained by one imaging device often cannot contain enough information for clinical purposes.At this time,the medical images obtained by multiple imaging devices need to be merged to be fully enriched Patient information,and the premise of image fusion is registration technology.This article takes the pre-radiation positioning as a scene to study the multimode medical image registration technology.In the radiotherapy planning software,the CT(Computer tomography)image of the patient taken before the plan is used as the reference image and the CBCT taken before the fractional treatment(Cone-Beam Computer tomography)The image is a floating image,and the registration error of 6degrees of freedom is obtained to correct the position of the patient and achieve the purpose of accurate radiotherapy.The main work of this thesis is as follows:Firstly,this paper proposes an improved mutual information method for registration similarity measurement.Mutual information is the most widely used similarity measure in multi-mode medical image registration.Image registration based on mutual information does not require pre-processing operations such as image segmentation and feature extraction in advance,nor does it need to assume the gray-scale relationship between images in advance.Then,the calculation method and process of mutual information of the image are deduced and analyzed from the calculation process of mutual information.The calculation area of normalized mutual information is improved,taking into account the spatial information between image pixels,so that the value of mutual information more accurately reflects the degree of registration.Secondly,this paper uses the improved Powell algorithm as the optimization strategy.The process of medical image registration is the optimization process of the objective function,that is,to find the spatial transformation parameters of the floating image aligned to the reference image when the mutual information value reaches the maximum.Aiming at the registration situation in this paper,Powell algorithm with strong local optimization ability is adopted.This algorithm makes full use of the property that the conjugate direction can accelerate the convergence speed.It can be used even when the derivative of the objective function is discontinuous.However,the algorithm has a high dependence on the initial point.The initial point position affects the convergence speed of the optimization process,and even affects the final registration result.In this paper,an improved algorithm is proposed to search for the initial point position,saving the initial registration Search time.Finally,based on the improvement of the above two key algorithms,a medical image registration system is designed and implemented,which can read,display and register medical images.The actual clinical data was used to verify the feasibility and stability of the system.