.

Recent developments in medical implant technology have seen a rise in devices designed for minimally invasive implantation in soft tissue. They allow to treat patients not considered for surgery, reduce the trauma caused by surgery, and thereby, reduce pain and improve the recovery time. For example, in the field of cardiology, transcatheter implants to treat aneurysms, vascular occlusion, and diseased heart valves are already available. However, the range of manipulations that can be performed during a minimally invasive surgery is limited. For example, in a transcatheter intervention, all instruments have to be flexible enough, and limited in diameter in order to pass through the vessels. This limits the ability to remove diseased tissue during the surgery, and to manipulate the tissue to fit closely around implant.  

In a transcatheter aortic valve implantation the diseased valve is not removed. Instead, the implant is placed inside the native calcified valve. This can cause problems with the sealing around the new valve, resulting in a backflow of blood, alongside the implant, when the valve is closed. The limitations of the transcatheter approach increase the importance of device selection, the implant's size, and the implantation position. Computer aided surgical planning can provide additional information about the 3-dimensional geometry of the anatomy, and the interaction between the implant and the native tissue prior to the surgery. Therefore, the aim of this thesis is to develop computer aided surgical planning tools to improve device implantation in soft tissue.

First, the aortic root of a population of transcatheter aortic valve implantation patients was characterised, based on the pre-interventional computed tomography images, to determine whether certain anatomical characteristics increased the risk of leakage after the implantation. A method was developed to calculate the calcification volume based on the images, and the geometry of the aortic root was characterised at two cross sections by multiple parameters such as diameter, area, etc. This investigation found that the patients suffering from moderate to severe aortic regurgitation post implantation, had a higher average calcification volume. However, there was a large overlap between the two patient groups. Whether an individual patient would develop severe aortic regurgitation could not be predicted based on the calcification volume.

Next, a method was developed to include the 3-dimensional anatomical shape of the aortic root in the evaluation of the optimal implant size. A segmentation algorithm was developed to automatically extract the aortic root and the aortic valve leaflets from the pre-interventional images. Based on the models of the aortic root, a parametric description was created using the population average, and the principal modes of variation. In turn, the parametric description was used to develop and train a classification method that could automatically determine the optimal size of the implant. The result was an algorithm that assigned the same size implant as was implanted to 96% of the patients in the training set. When applied to the patients with moderate to severe aortic regurgitation, 54% of them were assigned a different size than was implanted. Further research is required to assess whether the different size implants would improve the outcome of those patients.

In order to incorporate the interaction between the calcified native valve and the implant in the pre-interventional planning, a method for the patient specific simulation of the implantation was developed. The simulated stent was compared with post-intervention scans to validate the method. Based on the simulation result an algorithm was developed to estimate the post-interventional aortic regurgitation by detecting leakage paths between the modelled stent and the aortic root. The method was tested by simulating the implantation of 10 patients. The method could accurately predict the shape of the post-operative result, and the regurgitation estimation shows promising results that should be further validated on a larger patient set.

Finally, the shape model of the aortic root and the simulation method were combined to develop a framework for virtual testing of the fit of a device in the patient population. This method could be used to optimise the design of the next generation of implants. Calcifications were assigned to instances of the shape model based on the pre-interventional scans of the patient population. These models were used to simulate the implantation. Currently, the results are less accurate than the patient specific simulations. Further research is needed to improve this method in order to realise its potential.

In conclusion, this PhD thesis showed that computer aided surgical planning has the potential to improve the outcome for transcatheter aortic valve implantation patients by using the full potential of the 3-dimensional anatomical shape information, and by providing additional information about the interaction between the implant and the calcified aortic root. This will allow to make a more informed decision on the optimal size and position of the implant.