The team will develop a set of algorithms for use on high performance computers to analyze de-identified brain data from patients in order to better understand what electrical oscillations tell us about rapidly changing behavioral and pathological brain states.
When groups of synapses and neurons are active, they give rise to electrical oscillations that can be seen in the surrounding extracellular local field potential (LFP). These field potentials oscillations are far easier to record than the activity of each of the surrounding neurons, but their relationship to the precise activity of neurons is not simple or linear. The frequencies of these oscillations span a wide range: from slow delta sleep rhythms (0.3-4 Hz) to fast ripples (150-210 Hz), and each is coarsely correlated with a unique set of behavioral or pathological brain states. However, brain rhythms are not stationary and are in constant fluctuation: no two consecutive cycles are ever the same. Despite this continuous fluctuation in the amplitude, duration, slope, shape and microstructure of each oscillatory cycle, the majority of the neuroscience field continues to use standard frequency-domain methods based on traditional signal processing paradigms.
These traditional approaches rely on the average frequency and amplitude over a given period of time, preventing us from understanding how the brain computes and processes information on the true, instantaneous time-scale of neural activity. This project will use the instantaneous amplitude, frequency and slope of human oscillations to predict single neuron spike rate and timing, and will design novel algorithms to extract the instantaneous waveform shape of oscillatory cycles and use this shape to predict single neuron spike rate and timing.