The advance in the neuroimaging technology has brought with itself many challenges. One key challenge is the identification of dynamic functional networks underlying the observed neural activity. The current imaging modalities reflect solely the local neural activity, and do not provide a measure of the interaction between electrical or magnetic activity recorded at different sites. In order to gain a better understanding of how the brain processes information, it is crucial to quantify these interactions across the brain. This research addresses this problem by developing a comprehensive signal processing framework to study the functionality of the brain as a complex system based on electroencephalogram (EEG) data for a better understanding of psychopathologies.
This research develops three major signal processing methods to improve the current understanding of the brain as an integrated information processing system. First, data reduction and source separation methods that use the sparsity of the EEG signals in the time-frequency plane are developed to reduce large amounts of data into a few neuronal sources.
Second, synchrony and information based functional connectivity measures are developed to quantify the time-varying dependence between different neuronal sources. Third, methods from graph theory are adopted to build network models based on the functional connectivity measures and to identify the neuronal pathways and modules. Finally, this framework is applied to the study of the brain related to different psychopathologies including schizophrenia and impulse control problems. Several educational programs are tightly integrated with these research activities to emphasize the societal benefits of engineering. In particular, the investigator is designing outreach activities for K-12 women students and developing an undergraduate signal processing course with a focus on neuroscience applications.
Normal 0 false false false EN-US X-NONE X-NONE Intellectual Merit: This project titled ‘CAREER: Integrated Research and Education in Functional Brain Networks’ developed three complementary signal processing approaches to quantify and understand the connectivity patterns in the human brain during cognition. The first approach focused on quantifying pairwise relationships between brain regions from multichannel electroencephalogram (EEG) recordings. This approach developed a high resolution complex time-frequency distribution, named as Reduced Interference Rihaczek Distrbution (RID-Rihaczek), and a corresponding phase and phase synchrony estimation approach. This new distribution was shown to have better resolution, improved robustness to noise and less bias compared to existing methods such as wavelet based synchrony estimators (see Figure 1). This bivariate phase synchrony index was then extended to the multivariate case by introducing the hyperspherical phase synchrony measure. These measures of synchrony were applied to 64-channel EEG recordings during a study of error-related negativity. Error-Related Negativity (ERN) is an event-related potential (ERP) peak that occurs 50-100ms after the commission of a speeded motor response that the subject immediately realizes to be an error. Previous analysis has shown that ERN is dominated by partial phase-locking of intermittent theta band (3-7 Hz) EEG activity. The primary neural generator of ERN is the anterior cingulate cortex (ACC) and ERN exhibits itself through increased coordination between ACC and lateral prefrontal cortex (LPFC ) and the motor cortex. In our project, we showed increased phase synchrony between the frontal central electrodes and the right lateral area in accordance with this hypothesis. This increased phase synchronization was shown to be significant with respect to the correct response (see Figure 2). The second approach developed as part of this project focused on studying the brain network from a graph theoretic perspective. Using the bivariate phase synchrony value between different brain regions, a graph with nodes corresponding to the different regions and weighted edges corresponding to the pairwise synchrony values was constructed. New weighted graph theoretic measures including the weighted path length, weighted clustering coefficient and weighted small-world parameters were developed (Figure 3). A new hierarchical clustering algorithm based on spectral clustering was also developed to identify functionally common clusters in these networks across multiple subjects (see Figure 4). The final approach developed as part of this project aims at quantifying the causal relationships between different brain regions or neuronal populations. This approach aimed at differentiating between direct and indirect connections in the brain by quantifying the pairwise relationships through a directed information measure. Directed information (DI) is a measure of causality that quantifies the effect of the past samples of a given time series on the current sample of another series. In our work, we developed computationally efficient ways of estimating DI from limited data samples and showed this to be a more effective way of quantifying causal relationships compared to existing methods such as Granger causality especially in the case of nonlinear interactions. We have also extended this method for network inference and community finding applications and introduced a new directed weighted community detection algorithm. This hierarchical algorithm was applied to EEG data to discover clusters for ERN activity. Similar to our work using phase synchrony, we found distinct clusters corresponding to the motor cortex and LPFC. Broader Impact: As part of this project, one underrepresented minority (Hispanic) and one female Ph.D. student were supported. The results of this project in the area of functional connectivity brain networks were implemented as part of a summer outreach program to high school students and a bioinstrumentation lab experiment was designed to introduce students to the biomedical applications of engineering. These experiments and demonstrations were also used as part of Women in Engineering summer program. The scientific findings from this project were disseminated through eight journal publications and twenty eight conference papers. The findings from this project were also presented at interdisciplinary workshops and conferences, including special session presentations at 2011 Asilomar Conference on Signals, Systems and Computers, 2010 IEEE International Conference on Acoustics, Speech and Signal Processing, 2009 Society for Psychophysiological Research Annual Meeting and 2009 International Congress on Event-Related Potentials of the Brain.
