Optimization and Fiber-Centered Prediction of Functional Network ROIs

Study of functional and structural brain networks via fMRI and DTI data has received significant interest recently. A fundamental and challenging problem to identify a specific brain networks is how to localize the best possible regions of interests (ROIs). In this paper, we firstly propose a new approach to quantitatively describe fiber bundle and measure the similarity of two fiber bundles. Then we present a novel framework to optimize the shape of ROIs by maximizing fiber bundles similarity cross subjects and predict brain network ROIs in individual brain based only on DTI data. Our experimental results show that optimized ROIs have significantly improved consistency in structural profiles across subjects and demonstrated that fiber bundle description model derived from DTI data is a good predictor of functional ROIs. This capability of accurately predicting brain network ROIs would open up many applications in brain imaging that rely on identification of functional ROIs.