The Applation Research On Surface Reconstructon Methods For 3D Medical Ultrasound Images

Ultrasonic imaging is widely used in the clinic diagnosis for its advantages of being non-invasive, no-pain, real-time and convenient. Now, more than 25% medical imaging studies is related with ultrasonic imaging. There are many clinical application limitations for conventional 2D ultrasound: Physicians need to imagine the 3D organ just from many 2D ultrasound images when diagnosing, it can lead to low efficiency and errors; 2D ultrasound images only depict a perspective organ, it is difficult to obtain the most optimal imaging plane for the diagnosis. The questions above can be solved by the 3D ultrasound technology. The 3D ultrasound reconstruction technology and 3D visualization technology can reconstruct and display the models of organs, parenchyma and the pathological changes, which supply an intuitionistic vision to the doctors and improve the precision. There are two main methods in the 3D ultrasound visualization technology, one is surface rendering algorithm which is also called surface reconstruction, the other is volume rendering algorithm. Volume rendering can reflect the real structure of the body. But it is with large computing and long rendering time, even though using high-powered computer it can not meet the requirement for real-time interactive analysis. So in recent, the surface rendering is still the main method. This paper realized an algorithm: reconstruct the surface of the object based on the origin cross sections, which is especially suit to the sensor localization freehand arm 3D US system.The contour points of the object should be first obtained through segmenting and interpolating, and then rendered to reconstruct the surface. The paper mostly researches the segmenting and interpolating. The research content contains two main parts. One part is to improve the T-Snake model further and use this method to segment the contour points of the object on the cross sections. Seen from the experiment, the result is good. The other part is to improve the shape based interpolation directed by maximal discs and use it to interpolate the contour points of object among the cross sections. The cross sections are non-parallel, this interpolation method can correct the interpolating direction exactly in time, which ensures the good reconstruction quality. At the same time, we use a little computing skills in the interpolating the mid-point, so the time is reduced.