Construction of 3D volumetric shape model from 2D US images

Abstract

The current medical and engineering technologies help us in various fields of surgeries such as surgical simulations, preoperative plannings, intraoperative supports, and diagnostic explanations. In preoperative plannings, the simulation model of the target organ is necessary, and it is usually made from fine images that are obtained by some imaging devices such as CT (Computed Tomography) or MRI (Magnetic Resonance Imaging). However, these devices are too big to use during the surgery although there is a MRI device that can be used during the operation. On the other hand, the device of Ultrasonography (US Echo) is so handy that it can be easily used during the operation without radiation or magnetic effect; however, US images are not usually used for the construction of 3D model because they have so many noises and are not as clear as CT or MRI. Therefore, we have tried to construct 3D volumetric shape model from 2D US images by using image processing technique. Our concrete target organ and surgery are the aorta and Annuloaortic Ectasia (AAE), respectively. In order to generate the 3D volumetric aorta model from 2D US images, two types of US images are used: the horizontal and the vertical cross section images (See Fig.1 (a) and (b)). The horizontal cross section image is obtained from the top view, which can see the circle of the aorta wall, while the vertical cross section image is obtained from the side view, which can see the aorta wall and its valve. First of all, the binary image is made from the original image by binarization process, which is one of image processing techniques. The binary image, however, has too many noises to process so that the image is thinned by contour extraction process in order to extract the aorta contour. Then, the wall and the valve of the aorta are estimated with the information of the wall, which is obtained by changing the threshold of binarization to the original images. In order to generate 2D clear image of the aorta, the above processes are applied to both of the horizontal and the vertical images. Finally, 3D volumetric shape model can be constructed by inserting the generated 2D horizontal image into the corresponding position of the 2D vertical image. In this inserting process, the horizontal image should be shrunk or expanded according to the size of the corresponding position of the vertical image.