|Abstract:|| Nowadays ventricular assist devices play an important role in the treatment of terminal heart failure. While the devices themselves have been widely studied there are no studies of patient-specific numerical simulation in this context. This could be explained by the fact that the presence of the device induces metallic artifacts and noise in the acquired images so that conventional segmentation techniques fail. The aim of our work is to propose a robust framework for the segmentation of medical images of poor quality, the generation of high quality meshes and for the patient-specific analysis of the collected data via fluid-structure interaction (FSI) numerical simulations. First images are processed using histogram adjustment, histogram equalization, and gradient anisotropic diffusion filter. The watershed algorithm is then applied and the result is refined by the use of morphological operators. Then our framework allows the generation of two conforming meshes, one for the arterial lumen and the other for the arterial wall, ready for FSI simulations.
We also describe the numerical model and methods used to perform FSI simulations. Final results performed on two patients demonstrate the ability of our methods: the whole strategy results suitable, robust, and accurate for patient-specific data.|