While the specific functions of sleep are still being debated, sleep quality certainly impairs memory, cognition, vigilance, and motor skills to varying degrees. Animal experiments can uncover new knowledge about sleep and its interactions with aging, brain injury, epilepsy, and other conditions. But this presumes the ability to discriminate and track the brain's instantaneous vigilance state (e.g., REM sleep, non-REM sleep, Wake) continuously across the sleep-wake cycle. We are using discrete Bayesian models to map noisy continuous-valued measurements (EEG, EMG, motion) onto recognized vigilance states. We have set up a system for high quality video-EEG monitoring and analysis in mice that classifies vigilance state with over 90% accuracy. In addition, we use minimally intrusive sensory stimulation to study the consequences of selective sleep perturbation. We expect these modeling and technology development efforts to facilitate investigation of sleep as well as the study of sleep-seizure interactions.
Epilepsy is a devastating brain disorder characterized by spontaneously recurring seizures. A seizure is loosely described as a paroxysmal neural discharge in the brain that can impair consciousness. Epilepsy affects over 1% of the world's population, and roughly a third of all epileptic individuals suffer from seizures that are resistant to drugs or surgical intervention. We would like to know why seizures occur when they do, and how they might be stopped. Sleep and seizures interact in complex ways: seizures interrupt sleep, and most forms of treatment can themselves alter sleep or impair vigilance and cognitive performance; furthermore, sleep quality can influence seizure generation. Our research is directed toward analyzing the sleep-seizure interplay through physiological signal analysis and computational modeling in humans and in rodent models of epilepsy, with the end goal of optimizing treatment dose and timing to achieve seizure control without compromising sleep quality.
Injuries to the central nervous system can impair the ability to perform everyday motor tasks. Brain-machine interfaces (BMIs) have been developed to decode brain signals into control commands for prosthetic devices. We have set up our own BMI to be driven by the sensorimotor rhythm ("mu" rhythm) of the EEG. We have performed studies on control subjects with IRB approval, and found that detection of EEG changes in close temporal proximity to left/right hand movement is feasible. We are investigating ways to extract graded event-related potentials, or GERPs, from the EEG that reflect varying degrees of intent to enable smooth BMI control. Beyond their use as assistive devices, we believe that BMIs can up-modulate neuroplastic change (brain reorganization) and facilitate motor recovery. Repetitive exercise can induce beneficial neuroplastic change in the motor cortex, and electrical stimulation of peripheral nerves or muscles can augment this recovery. Neuroplastic change could be enhanced further by synchronizing peripheral afferent nerve stimulation (PNS) with the intent-to-move. To this end, we propose to use a BMI to trigger PNS in closed-loop. With IRB approval, we have started screening patients with motor incomplete spinal cord injury for inclusion using a brief EEG session. We expect this study to generate valuable information on the relative contributions of different modes of sensory feedback and timing-dependent neuroplasticity in patients undergoing rehabilitative treatment for motor disabilities.