Evaluation and optimisation of a catheter simulation for endovascular navigation

Abstract

Catheter navigation is a key issue of contemporary endovascular intervention. The navigation still requires a high level of training and experience to be successful and nevertheless comes with numerous intricacies like radiation exposure and negative side effects from contrast agents. To overcome the steep learning curve and error susceptibility, as well as other issues, a catheter navigation approach was developed. The alleged solution is to simulate a catheter path, using the patients vessel system as basis, and then compare its curvature to the measured curvature of the actual catheter inside the patient in order to localize its position in the human body. For this master thesis, the given task is to detect the accuracy of a catheter simulation program and optimise its implemented parameters in order to improve the catheter simulation. In order to do this, a segmentation network is developed, which extracts and processes the image data needed for the simulation. Subsequently, an evaluation network is developed to determine the accuracy of a set simulation module by comparing the catheter simulation to the real catheter that moves through the vessel. Finally, the parameters embedded into the simulation algorithm are strategically changed in order to lower the error value of the simulation. The aim is to determine the accuracy of the simulation module in relation to reality and thus assess its suitability for use in the medical environment. If the simulation approach is proven to be close to reality, the optimal parameter setting must be modified with regards to the computational strain and accuracy.