A framework of image processing for image-guided procedures
MetadataShow full item record
Contrary to the cases of open surgery, during minimally invasive procedures, surgeons have no direct access or very limited access to the operative site and can only visualize the site and the focal region with the aid of imaging tools. During patient follow-up and the surgical planning stage, multiple complementary images and patient-specific models need to be aligned with each other for better interpretability. Image segmentation and deformable image registration are often needed for this step since the tissues normally deform in different images. Traditionally, image segmentation and registration are separate research topics. However, they are closed related. Segmentation accuracy affects feature based registration accuracy and a good registration is able to improve the speed and accuracy of image segmentation. In this thesis, we present a coherent and consistent framework for image processing. The process of the proposed approach consists of two main modules. In the first module, we performed preprocessing tasks on raw MR images to extract the blood vessels centerlines from these images. Next, bifurcation points were calculated as the intersections of the extracted centerlines. Vessel structures and bifurcation points provide input information for further operations such as image segmentation and image registration. Then, we propose an innovative approach to automating the initialization process of the liver segmentation of magnetic resonance images. The seed points, which are needed to initialize the segmentation process, are extracted and classified by using affine invariant moments and artificial neural network. To complete this stage, we proposed robust and fast approaches for MR image registration for deformable tissues. The proposed registration methods work with soft homogenous tissue with small and large rotation shift. In the second module, we performed surface to surface registration of MR and endoscopic images, which will provide the surgeon with better 3D context of the surgical site in minimally invasive procedures. In this step, we project a gridline light pattern onto the surgical site and then use a stereo endoscope to acquire two stereo images. The major steps in the surface reconstruction process include 1) applying an automatic method of detecting region of interest, 2) applying an image intensity correction algorithm, and 3) applying a novel automatic method to match the intersection points of the gridline pattern. We have validated our proposed framework technique by comparing the methods with existing techniques of similar scope. Our experiment results show that our methods outperform the existing methods regarding correctness and efficiency.