1. OPTIMIZATION OF FUNCTIONAL BRAIN ROIS VIA MAXIMIZATION OF CONSISTENCY OF STRUCTURAL CONNECTIVITY PROFILES Dajiang Zhu Computer Science Department The University of Georgia zhu@cs.uga.edu Introduction Segregation and integration are two general principles of the brain’s functional architecture; therefore brain network analysis is of significant importance in understanding brain function. Critical to brain network analysis and construction is the identification of reliable, reproducible and accurate network nodes, or Regions of Interest (ROIs). In this paper, based on functional ROIs derived from task-based fMRI, we propose a novel framework to optimize the location and size of the ROI which ensure the difference of structural connectivity profiles among a group of subjects is minimized. In order to facilitate the optimization procedure, we present a new approach to describe and measure the fiber bundle similarity quantitatively within and across subjects. This framework has been extensively evaluated and our experimental results suggest the promise of our approaches. This capability of accurately localizing brain network ROIs would open up many applications in brain imaging that rely on identification of functional ROIs. Motivation Identification of reliable, reproducible and accurate ROIs  The boundaries between cortical brain regions are unclear  The individual variability of cortical anatomy, connection and function is tremendous  The properties of ROIs are highly nonlinear Related Work Manual labeling by experts based on their domain knowledge[1] Cluster ROIs based on multivariate methods that, for example, calculate the amount of variance the ROIs account for or decompose the data into statistically independent ROIs[2] Template warping based on image registration[3] Using task-based fMRI paradigms to identify activated brain regions as ROIs Approach Contributions We presented a novel framework for the optimization of both locations and sizes of ROIs via maximization of group- wise consistency of ROIs’ structural connectivity profiles We proposed a novel approach for quantitative measurement of the similarity of ROIs’ structural connectivity profiles by projecting the fiber curves onto a standard spherical space References 1. Bharat B. Biswal,“Toward discovery science of human brain function,” PNAS, vol. 107 no. 10 4734-4739, 2010 2. Yufeng Zang, et al., “Regional homogeneity approach to fMRI data analysis,” NeuroImage, 22(1): p. 394-400, 2004. 3. David C. Van Essen, et al., “Surface-Based and Probabilistic Atlases of Primate Cerebral Cortex,” Neuron, 56, 2007. Acknowledgments I am heartily thankful to my supervisor, Dr. Tianming Liu, for his encouragement, guidance and support from the initial to the final level. Fig. 1. Illustration of structural and functional connectivity changes when the location of a ROI is changed slightly. (a) ROI location moves from the green to the red bubble. (b) Structural and functional (fMRI signal) profiles before the movement. (c) Structural and functional profiles after the movement. Fig. 2. The optimization scheme. (a) A group of subjects with initial locations and sizes of ROIs indicated by yellow circles. (b) A group of fiber bundle candidates for each ROI. (c) Trace-maps corresponding to each fiber bundle. (d) Distance matrices of different track-maps for each subject. (e) Typical fiber bundles for each subject after AP clustering. (f) The optimized fiber bundle (locations and sizes). (g) The movement from initial location (green) to the optimized location (red). (1) Extracting fiber bundles from different locations and sizes close to the initial ROI. (2) Transforming fiber bundles to trace-maps. (3) Calculating the similarity of different trace-maps within subjects. (4) Using AP clustering to find the typical fiber bundles for each subject. (5) Finding the group of fiber bundles which make the group variance the least. (6) Finding the optimized location and size of the ROI. Fig. 3. (a) Calculation of the principal direction for one segment of each fiber. (b) Each segment could be represented by a series of vectors. (c) After translation to the origin of a global coordinate system, each vector shoots to a unit sphere whose center is the origin. (d) Two examples of fiber bundles and their trace-maps. The top row is a U-shape fiber bundle example and the bottom row is a line-shape one. For both cases, the left are fiber bundles and right are their trace-map representations. Fig. 4. (a), (b) and (c), (d) are two pairs of similar fiber bundles. (e)-(h) are their respective trace- maps. Fig. 5. Illustration of comparison of two trace- maps. The point densities in red circles are compared. Fig. 6. Optimization results of 15 subjects and 16 ROIs. Each panel represents one ROI from 15 subjects. The left and right columns are the results before and after optimization. Some ROIs are highlighted. Fig. 7. Comparison of group variance before and after optimization. (a) Comparison of 15 subjects. (b) and (c) Comparison of sub-group after we randomly split all subjects to two groups.. Fig. 8. The movement of one ROI after optimization. The green ball is the original location, and the red ball is the new location. Yellow arrow indicates the direction of the movement.