Realistic forward and inverse model mesh generation for rapid three-dimensional thoracic electrical impedance imaging.

One of the most promising clinical applications of Electrical Impedance Tomography (EIT) is real-time monitoring of lung function in ambulatory or ICU due to the rapid, non-invasive and non-ionizing nature of the measurements. However, to move this modality into routine clinical use will, as a minimum, require the development of realistic and computationally efficient forward and inverse meshes of the thorax and the lungs alongside mechanisms to extract quantitative information from the resulting reconstructed images. The latter will allow for reduction of artefacts and better localization of conductivity changes within the image domain. This research aims to contribute towards this goal, by introducing a pipeline for the generation of anatomically accurate meshes for EIT forward and inverse models. We achieve this by the segmentation of realistic volumetric data from thoracic CT volumes, and subsequent tessellation. Mesh quality is assessed in terms of aspect ratio, dihedral and face angles. Moreover, the generated meshes are fused with currently available EIT software, with a novel electrode placement method, to show the practical application of the generated meshes. Results show that anatomically constrained unstructured meshes can be generated, conforming to realistic anatomical geometry, and performing well in EIT numerical computations. Such realistic computational models will further enhance the performance of EIT reconstruction algorithms, thus offering significant benefits to clinical EIT lung imaging.