2D-3D Medical Registration Based On Computer Aided Surgery System

Computer Aided Surgery System usually use 2D image sequence got before surgery to establish 3D model of anatomical structure by making use of 3D reconstruction and visualization method. Currently, doctors can use Computer Aided Surgery System to design surgical approach very well before surgery, while they can hardly combine images got in and before surgery to modify or even redesign the surgical approach.Owing to its advantage of real-time, ultrasound imaging is widely applied in clinical.However,its shortcomings are also very obvious, such as:existing some speckle noises, limited scanning range, precision decreases as scanning range increases. If the system can complement the advantages of ultrasound images and CT image sequence, it will be possible to track the organs in the surgery.3D ultrasound guided navigation is being widely studied and applied abroad, while this method is not so popular in domestic application. According to this situation, this dissertation discusses and implements a 2D-3D medical image registration system based on Computer Aided Surgery System which can provide doctors with intuitive help.This dissertation focus on some important algorithms and their implementations which include ultrasound image simulation, ultrasound image filtering and image registration.In ultrasound image simulation, we introduce a ultrasound image acquisition model as well as a ultrasound image noise-adding model. It has three steps:ultrasound image sampling, noise adding and interpolation. This simulation algorithm makes it possible to provide filtering algorithm evaluation with objective criteria.In ultrasound image filtering, we introduce two kinds of algorithms:one based on mathematical morphology method, the other based on median filter method. The former one focuses on a algorithm called Open-By-Reconstruction by combining two basic operators named dilation and erosion. This algorithm can preserve image's scale values by and large while eliminate some highlight points. The later one consists of three variants:standard median filter, multi-stage median filter and adaptive weighted median filter.In image registration, we introduce two optimization algorithms:the Powell Algorithm and the Particle Swarm Algorithm. These two algorithms can find the best solution in search space spontaneously to get all the parameters in image registration.At last, according to these parameters, we can calculate the coordinate transformation function between the CT space and the word space.