Reading Your Own Mind: Dynamic Visualization of Real-Time Neural Signals

Brain Computer Interfaces: BCI) systems which allow humans to control external devices directly from brain activity, are becoming increasingly popular due to dramatic advances in the ability to both capture and interpret brain signals. Further advancing BCI systems is a compelling goal both because of the neurophysiology insights gained from deriving a control signal from brain activity and because of the potential for direct brain control of external devices in applications such as brain injury recovery, human prosthetics, and robotics. The dynamic and adaptive nature of the brain makes it difficult to create classifiers or control systems that will remain effective over time. However it is precisely these qualities that offer the potential to use feedback to build on simple features and create complex control features that are robust over time. This dissertation presents work that addresses these opportunities for the specific case of Electrocorticography: ECoG) recordings from clinical epilepsy patients. First, queued patient tasks were used to explore the predictive nature of both local and global features of the ECoG signal. Second, an algorithm was developed and tested for estimating the most informative features from naive observations of ECoG signal. Third, a software system was built and tested that facilitates real-time visualizations of ECoG signal patients and allows ECoG epilepsy patients to engage in an interactive BCI control feature screening process