Introduction: Image based navigation is an important tool for the orientation during minimally invasive surgery. The registration of the preoperative image data for orthopaedic surgery is nowadays mostly based on anatomical landmarks and bone surface points. Only the bony structures can be registered reliably with these systems. Since the location of the patient has changed after acquiring the preoperative dataset, soft tissue such as vessels, sinews or muscles might be disarranged. For performing the surgery exactly, the current position of soft tissue can be very important. Our aim is to segment blood vessels in intraoperative 3D ultrasound datasets in order to support the surgeon's orientation during surgery. After registration of intraoperative 3D ultrasound data with preoperative data [1], we can superimpose the segmented blood vessels at their current position with the registered bone structures. Most of the current segmentation methods like thresholding or region growing fail in segmenting blood vessels in ultrasound images because of two main reasons: The contour of a vessel is non-continuous due to shadowing effects and speckle, and the gray value is not constant all over the vessel. By modifying the convential region growing algorithm with an assumed vessel model, the proposed method compensates most of the artifacts in the ultrasound image data.
展开▼
