The Research Of Medical Image Three-dimensional Reconstruction

1. IntroductionIn medical diagnoses, observing patient's a set of CT image is the conventional mode for doctor to diagnose illness. Now the image from medical computed tomography although can reflect the dislocation internal organized information, it can't get the three-dimensional organization pattern and the interrelate information among organizations. But it's too important for make sure the organization quality. Only depending on CT dislocation image information to determine the pathogeny's location,size,geometric shape and the spatial relation with around organization is quite difficult, so we need a try and true instrument to complete the three-dimensional reconstruction and three-dimensional display. The technology of CT Image three-dimensional reconstruction is the effective instrument for doctor to analysis the pathology and around organization, which greatly improves the medical diagnostic accuracy.2. Research ContentThe purpose of this study is the research of CT image three-dimensional reconstruction. The original image is based on DICOM3.0 standard, after image format conversion, image preprocessing, image segmentation, we proceed three-dimensional reconstruction. The primary research content includes five parts.(1) The study of DICOM 3.0 standardThe DICOM standard is specific for medical image storage and transmission, which used to specify the image format and relevant information. Because its image format is different from some common image format ,such as JPEG, TIF, BMP, some image processing software can't display, process and convert it, we need to convert DICOM image to common image format. The paper introduces the DICOM file standard, file structure and how to convert DICOM image to BMP image. We realize to read DICOM image and convert DICOM image to BMPimage in software platform.(2) The CT image preprocessingThe medical image is produced by image equipment, because of the influence of medical equipment and output condition, the original image could appear image quality degeneration, even artifact. Therefore, medical image three-dimensional reconstruction, the primacy thing is preceed pretreatment towards medical image. The paper introduces the technology of CT image preprocessing in detail, including image filter and interpolation. In image filter algorithm, I introduce the neighborhood filter and medium filter. In image interpolation algorithm, I introduce the linear weighting interpolation and tri-linear interpolation. Though analyses and compare these algorithm's experimental result, I choose medium filter and tri-linear interpolation.(3) CT image segmentationCT image segmentation is the foundation of three-dimensional reconstruction, the effect of segmentation will directly influence the accuracy of three-dimensional reconstruction mode. Segmentation can help doctor extract interested object, and help doctor qualitative and quantitative analyses the pathology organization, thereby advanced the medical diagnostic accuracy.According to the difference of segmentation algorithm adaptability, image segmentation has two methods: one is edge segmentation, which uses the region heterogeneity to mark off the region boundary. The other method is based on regional segmentation, usually use the same intradistrict uniformity to identify the image different region. The paper introduces both of these segmentation, in the algorithm of edge detection, I introduce some edge detection operators, and emulate them to compare with each other. In the region segmentation algorithm, I study the threshold segmentation and region growing segmentation, and improve the region growing segmentation, use the threshold segmentation and edge detection to region growing segmentation. I do experiment towards these algorithm in computer platform, the result indicates that the algorithm can get the satisfied results, charging off the headstock, overcoming the shortage of traditional region growing algorithm, which can segment cerebral cranium image quickly andeffectively.(4) CT image three-dimensional reconstructionMedical image three-dimensional reconstruction has two methods: surface reconstruction and volume reconstruction. The surface reconstruction construct middle geometrical pixel in three-dimensional space data, and use conventional computer graphic technology to realize surface reconstruction. Volume reconstruction directly converts three-dimensional space's discrete data to two-dimensional image, instead of producing geometrical pixel, the substance of which is resampling and image synthesis. The paper introduced some classical tri-dimensional reconstruction, analyzing and estimating them from three facts: rendering effect,rendering velocity and algorithm feature. Because of the limit of hardware condition, surface rendering's interactivity is greatly precede volume rendering, we emphases study CT brain image surface rendering, deeply study the classical algorithm–Marching Cubes algorithm. Because MC algorithm is sequential measuring each pixel, by research testified, when abstracting a isosurface, it costs 80%~90% time to detecting null pixel, improving MC algorithm hinge on decreasing detect null pixel, finding out pixel crossed to isosurface as soon as possible. Based on this considering, we adopt an algorithm which based on segmentation image, which use the segmentation image as input image. We can select optima segmentation method based on image characteristic, using the result of segmentation to construct isosurface. The paper adopts improved region segmentation image, at the same time of surface tracking, we adopt region growing cube detection method, which reduces to detect null body. Last we do experiment on computer platform, the result declares that the algorithm reduces the operative capacity and advances the reconstructed efficiency.(5) The design of medical image three–dimensional reconstructionThe system software is in Window XP system, use Visual C++ 6.0 to develop the medical image three–dimensional reconstruction system, which realizes the CT image three-dimensional surface reconstruction, has a widespread clinical application value. The system realizes the conversion of DICOM to BMP. In the image preprocessing, it realizes the neighborhood filter,medium filter andtri–linear interpolation; In segmentation algorithm, it realizes the threshold,region growing segmentation and advanced region growing segmentation; In the surface rendering algorithm, it realizes standard MC algorithm, as well as based on the improved region growing segmentation and segmental MC algorithm adjoint surface rendering. It also can do some interactive operation, such as zoom,revolute and shift the three–dimensional model.3. ConclusionThe paper primary studies the CT image three–dimensional reconstruction, and deeply introduced the key technology. Incorporation the newly success and trend of development in this field, combine the advanced region growing segmentation algorithm and advanced MC algorithm, use Visual C++6.0 to design a three–dimensional reconstruction, and realize the CT image tri-dimensional reconstruction system.