Research On Automatic Extraction And Reconstruction Of Meniscus In Magnetic Resonance Image

Meniscal injury is a common knee joint desease, and magnetic resonance imaging is an effective imaging method to detect meniscus lesions. While the magnetic resonance images are often of a large number, and always interfered by noise information. Doctor need to diagnose the disease by observing images piece of piece, and the result always affected by subjective factor, medical experience, imaging quality and some other enviroment factor. Computer technology, as an aided method, has been applied in clinical diagnosis in recent years. In this paper, for the computer-aided diagnosis of meniscus lesions, in order to solve the objects extraction problem in MRI sequences, an automatic target detection and extraction method based on thresholding and shape constraint is proposed to detect and extract the meniscus target in magnetic resonances images, and it achieved an autometic imaging sequence division process. At the same time, for the 3D displa y problem in the aided diagnosis, a meidcal image processing software based volume rendering and volume reconstruction methods and a Matching Cubes based surface rendering method are proposed to realized the meniscus reconstruction, on the basis of the extraction result. The research in this paper can be useful for doctors' diagnosis and computer aided diagnosis. At the same time it can help build teaching demonstration, operate simulation and 3D printing system.The research in this paper use the clinical medical knowledge as a basic guide and based on some classic computer imaging processing technology to bulid a meniscus automatic extraction system. And the 3D reconstruction is based on special software and reconstruction algorithm to achieve the extrudin g of the extraction result and validate the pratical of the research. The accuray rate of the extraction is 84.9% and the recall rate is 86.5%, and the accuracy rate for typical triangle shape meniscus is 88%. The algorithm efficiency increased by one times due to the improved parameter sets. And the region of interest positioning error problem and wrong extraction of articular cartilage problem are solved.Compare with the existing method, this system is more targeted to meniscus problem. It can achieve sequence batch, has a higher efficiency, easier to realize and has better extensibility. Meanwhile, a set of scoring regulars are formulated in the evaluation of the result of 3D reconstruction in order to judge the structual integrity, observability and operability of the 3D model. It's a helpful exploration for building a unified evaluation system to judge and compare the reconstruction effect by different ways instead of subjective judgement only.