Brain modeling guided by direct volume rendering

In the planning of neurosurgery, it is often required to build brain model from medical imaging data. It is hard for traditional brain modeling based on image segmentation to provide high quality models automatically. Direct volume rendering (DVR) is a powerful tool to disclose inner structures of volume data, but it does not produce explicit geometric objects. This causes it hard to use DVR in some applications based on virtual reality (VR) or mixed reality (MR). In this paper, we propose a new method to build brain model under the guidance of DVR, where a volume of brain imaging data is first visualized by DVR using properly set opacities and colors, and then a series of iso-surfaces that obtain their opacities and colors from DVR are extracted to build the brain model. Compared with brain modeling based on image segmentation, our method achieves higher quality of models by increasing accuracy and enriching detail of brain structures. Moreover, our method has the flexibility to explore brain structures interactively.