A Simple and Efficient Cylinder Imposter Approach to Visualize DTI Fiber Tracts

Abstract

The human brain can be divided into two tissue categories, namely: gray matter that maybe associated with cognitive, motor, emotion, and visual processing, and white matter that facilitates neuronal communication between gray matter regions. To better understand the organization of white matter connections in the brain, white matter fiber tracts derived from a diffusion tensor image scan is estimated and visualized by publically available software toolsets. In general, one white matter fiber tract is visualized as a thin 3D cylinder, however this approach has many computational limitations, especially when trying to visualize thousands of fiber tracts that have varying size and length. To overcome this limitation, a very simple and efficient imposter approach is proposed, commonly used in the computer graphics community, that exploits the programmable pipeline architecture found in GPU-based parallel processing systems. Using 10,000 fiber tracts derived from a real DTI scan, we show the rendering speed of our imposter approach is 50% times faster, and requires 900% less memory, when compared visualization approach that uses 3D cylinders.