Intrinsic Brain Geometry Revealed by Dimensionality Reduction
Immersive 3D visualization of Whole-Brain Tractography
This video displays a 3D interactive immersive environment for visualizing tractography from a single subject derived with standard diffusion tensor imaging (DTI) and high angular resolution diffusion imaging (HARDI). The visualization in CAVE2™ is sensitive to the user's point of view through the use of head-tracking 3D glasses. Movement through the virtual brain is controlled by the user with a modified Playstation Move controller. By visualizing tractography data in CAVE2, differences between DTI and HARDI in reconstructed fiber density and fiber geometry are easily appreciated.
CAVE2 is approximately 24 feet in diameter and 8 feet tall, and consists of 72 near-seamless passive stereo off-axis-optimized 3D LCD panels, a 36-node high-performance computer cluster, a 20-speaker surround audio system, a 10-camera optical tracking system and a 100-Gigabit/second connection to the outside world. CAVE2 provides users with a 320-degree panoramic environment for displaying information at 37 Megapixels in 3D or 74 Megapixels in 2D with a horizontal visual acuity of 20/20.
Path Length Associated Community Estimation (PLACE)
Path length associated community estimation (PLACE) is a comprehensive framework for studying node-level community structure. Instead of the well-known Q modularity metric, PLACE utilizes a novel metric, ΨPL , which measures the difference between intercommunity versus intracommunity path lengths PLACE consists of the following: (1) extracting community structure using top-down hierarchical binary trees, where a branch at each bifurcation denotes a collection of nodes that form a community at that level, (2) constructing and assessing mean group community structure, and (3) detecting node-level changes in community between groups.. In this video, we see how PLACE iteratively creates community structures over 4 stages.
Download the MATLAB code for PLACE here.
The Brain IsoMap (BRISOMAP)
The Brain Isomap is an innovative technique for conceptualizing and visualizing the intrinsic geometry of brain connectomes. Using a non-linear dimensionality reduction techniqe applied to the graph distance matrix, the Isomap represents the intrinsic geometry of the structural connectome in a d-dimensional Euclidean space. The video demonstrates that the optimal embedding was achieved with d=3 (3-dimensional space). The spatial distance of nodes in the Isomap reflect the graph distance of those nodes. Thus, highly connected nodes or rich-club nodes tend to be centrally located in the Isomap.
WTTW Coverage of the CAVE2 Connectome Visualization Project
Functional by Structural Hierarchical (FSH) Mapping
FSH Mapping is a novel multimodal integration method for creating connectomes from structural and functional imaging data. In brief, the method estimates the white matter structure underlying resting-state functional connectivity. FSH assumes that the resting-state functional connectivity between two regions can be modeled as an exponential decay function of the "modified" graph distance of the structural connectivity matrix subject to a utilization matrix that is estimated using simulated annealing.