My current focus is on developing finite-element models including the inherent anatomical variations. The idea is to mimic normal and pathological conditions of the middle ear and generate a large dataset of synthetic data. The dataset is used to train machine-learning algorithms to infer the middle-ear status from clinical data. There are a number of challenges regarding the development of a population of 3D geometries of middle-ears. The fine structural characteristics such as eardrum thickness are in the order of voxel dimensions of the μCT images, thus segmentation is usually done manually which is time-consuming and depends on the experience of the modeler.
The interfaces of different components are not usually clear and automatic segmentation algorithms do not work properly. Another challenge is importing the geometry into the finite-element software. Although Simpleware has powerful controlled meshing capabilities to generate compatible mesh, generating CAD geometries from meshes inside finite-element environment is always challenging. I look forward to seeing machine-learning methods implemented in Simpleware to perform segmentation, and also capabilities for developing parameterized geometries in the next versions of Simpleware.