Research On Surgical Navigation For Cardiovascular Intervention Using Intra-operative Ultrasound

Traditional cardiac intervention surgery to deploy the stent-graft at target position is achieved under the guidance of X-ray, which results in three distinct disadvantages: fluoroscopy exposes the clinicians, staff and patient to ionizing radiation; during the surgery, the patient must be injected with boluses of contrast agent, increasing the risk of iatrogenic renal injury; and fluoroscopy and angiography only provide 2D images yielding poor visualization in some regions of the aorta, giving rise to navigational limitations. Therefore, an alternative image-guidance procedure that mitigates these limitations is needed for cardiac intervention surgery. This thesis proposes a navigation system integrating the intra-operative ultrasound(US) image and preoperative CT image to guides the surgeon to position and deploy the stent-graft to the target, which makes the cardiac intervention surgery be performed without X-ray guidance possible. The combined navigation system was applied in stent-graft deployment for thoracic aorta and aortic valve, which were representative surgeries of the cardiac intervention surgery. The major work and contributions of this thesis include:1) An interactive segmentation based on region growing is proposed. It makes use of the GPU accelerated volume rendering proposed by this thesis enabling the surgeon to view the intermediate segmentation result intuitively to de termine constraint surface of next step rapidly and accurately. Compared to tr aditional manual segmentation, it could segment the target region more rapid and accurate.2) A GPU accelerated continuous max flow segmentation based on prior probability distribution is proposed to segment the contour of aortic valve of the short axis intra-operative US image accurately in real time. The extracted contour of aortic valve is used for the registration between the intra-operative US image and preoperative dynamic aortic model.3) The preoperative planning modules of stent-graft deployment for thoracic aorta and aortic valve are designed and implemented to determine the target position on the constructed cardiac model. An US probe calibration method is designed and implemented to integrate the US image into the navigation system.4) A two-stage registration between the intra-operative US image and preoperative aortic model is proposed to compensate the errors resulting from heart beating, respiration and body posture change.5) The phantom and animal studies were performed to validate the navigation system for stent-graft deployment of thoracic aorta and aortic valve.