Research Of The Segmentation And Registration For Liver Vessel Based On Multimodal Abdominal CT And MRI Images

With the improvement and development of medical image processing,biomedical technology,reverse engineering,3D reconstruction technology,and finite element analysis,computer-aided technology has been used in the pre-operative exercise planning,intraoperative assisted guidance and positioning,and postoperative recovery assessment During the treatment cycle.The liver is the largest organ of the human body except the skin,and the multiple and difficult to treat liver diseases also plague medical workers.Aiming at the difficulty of liver blood vessel segmentation,this paper proposes an automatic liver blood vessel segmentation algorithm with excellent segmentation performance.This algorithm combines the unified level set segmentation algorithm to optimize segmentation based on OTSU dynamic threshold segmentation.Five typical vascular segmentation algorithms were quantitatively evaluated using 20 cases of abdominal CT enhancement data from the 3Dircadb public dataset,and compared with the method of this paper.After the mixed image processing,and overcoming the difficulty of threshold setting and seed point location selection in the conventional region growth,the over-segmentation problem was avoided.The OTSU dynamic threshold can extract the approximate area of the blood vessel without setting a seed point.Optimize the segmentation results by unifying the level set.The level set is driven by the image edge information and the area information at the same time.Compared with the single image information driving,it can adapt to a larger balloon force,has better anti-edge leakage effect,and can solve the under-segmentation problem.Liver blood vessels can be more accurately segmented.The method was validated on a total of 20 liver datasets provided by 3Dircadb.The results showed that: compared with other methods,the method requires less interaction time and higher segmentation accuracy.Liver computer-assisted treatment plan based on preoperative electronic computed tomography(CT)and magnetic resonance imaging(MRI)data is an important auxiliary means for liver surgery and visualization of liver 3D anatomy.Preoperative computer-aided technology can help doctors pinpoint the location of blood vessels and tumors.Because CT images are more friendly for liver extraction modeling,MRI images for vascular tissue display are helpful for image segmentation.Therefore,after the 3D models of different information sources are established,coordinate system normalization processing is required.This paper proposes a non-rigid external feature registration method based on the portal vein centerline for the problems of liver blood vessel model and liver parenchyma model that do not correspond to the three-dimensional spatial reference,and the elasticity of liver tissue makes the resection a certain error.Specifically,the non-rigid iterative closest point(ICP)method and the multi-level B-spline transform method are used to extract the center line of the CT liver model and the MRI blood vessel model.The corresponding points of the two modal midline lines are determined in the iteration of the ICP algorithm The search of corresponding points is limited to a given search radius and vessel extension direction.Finally,a multi-level B-spline is used to approximate the non-rigid registration of the vessel surface.At the same time,L-BFGS-B algorithm is used to optimize the algorithm,which can effectively solve the problem that the registration algorithm takes up too much memory and reduces the efficiency of the algorithm,and can avoid falling into local extreme values.The algorithm has evaluated and evaluated 20 liver data in the 3Ddirect data set,and then applied it to the liver data of a patient.The results show that the relative position is relatively accurate,and no occurrence of blood vessels overflowing the liver surface has important significance for preoperative preview of liver surgery and medical education.