Published on: Oct 18, 2012
AustinMan and AustinWoman: High Fidelity, Reproducible, and Open-Source Electromagnetic Voxel Models
I presented my work on developing the AustinMan and AustinWoman Electromagnetic Voxel Models at the 34th Annual Meeting of the Bioelectromagnetics Society in Brisbane. These are two open-source anatomical human models that I began developing as an undergraduate when I joined the computational electromagnetics group at The University of Texas at Austin, where I am currently pursuing my PhD.
The group wanted to investigate the interaction of RF fields with the human body and was developing a novel electromagnetic simulator that solves the volume integral equation and is accelerated by an FFT-based algorithm. Of course, to do this investigation, we needed virtual human models. At the time, we looked around for available human models to use, but were disappointed with the options: Some models did not have very high resolutions while others were only available inside expensive software packages. Moreover, almost all of the models were lacking in documentation.
As you can imagine, it is quite a complicated job to develop a virtual human model, and along the way many assumptions and simplifications must be made; unfortunately, these assumptions that were made when the existing models were developed were not publicly available. As a result, we decided to develop our own models using the Visible Human dataset. To ensure that they had a high resolution, we used the color cross-sectional anatomical images. While identifying the different tissues in each slice, we carefully documented any assumptions that we made and which regions of the body we had difficulty with (perhaps due to tissues with similar colors or corrupt images).
After we finished our first model, we posted it online with the documentation and are continuing to update the models and make them available to researchers. Recently, we have posted the segmented slices (the raw data in developing such models) and the program we used to segment the slices; we adopted this open-source approach so that other researchers can also modify and improve the models. We currently have AustinMan v2.1, a whole body male model, and AustinWoman v1.0, a partial body female model, on the websites http://tiny.cc/AustinMan and http://tiny.cc/AustinWoman. I encourage BEMS members to visit these websites to learn more about the ‘black art’ of developing anatomical human models.
Looking forward, I hope that all groups developing such models embrace the transparency and open-source principles that we adopted when developing the AustinMan and AustinWoman models. I think these principles can enable progress in the accuracy and predictive power of anatomical human models. Otherwise, the trustworthiness of the models and the simulations that use them will remain highly questionable. I am currently working on various projects that use the AustinMan and AustinWoman models; these include quantifying errors in simulation results when lower-resolution models are used, benchmarking competing electromagnetic simulators, and analyzing antennas and communication channels for body-area networks.