Semi-automatic algorithm to build finite element numerical models of the human hearing system from Micro-CT data.

Abstract

Finite Element modeling has been an extended methodology to build numerical model to simulate the behavior of the hearing system. Due to the complexity of the system and the difficulties to reduce the uncertainties of the geometric data, they result in computationally expensive models, sometimes generic, representative of average geometries. It makes it difficult to validate the model with direct experimental data from the same specimen or to establish a patient-oriented modeling strategy. In the present paper, a first attempt to automatize the process of model building is made. The source information is geometrical information obtained from CT of the different elements that compose the system. Importing that data, we have designed the complete procedure to build a model including tympanic membrane, ossicular chain and cavities. The methodology includes the proper coupling of all the elements and the generation of the corresponding finite element model. The whole automatic procedure is not complete, as we need to make some human-assisted decisions; however, the model development time is reduced from 4 weeks to approximately 3 days. The goal of the modeling algorithm is to build a Finite Element Model with a limited computational cost. Several tasks as contour identification or model decimation are designed and integrated in order to follow a semi-automated process that allows generating a patient-oriented model.