Survey of Recent Neural Signal Processing Results
Emery N. Brown, M.D., Ph.D.
Institute for Medical Engineering and Science
Department of Brain and Cognitive Sciences, MIT
Department of Anesthesia, Critical Care and Pain Medicine
Massachusetts General Hospital/Harvard Medical School
In this presentation, I will discuss four problems in neuroscience data analysis in which we characterize/control the dynamic behavior of neural systems. Analyzing ensemble neural spiking activity is an important problem in neuroscience data analysis. We show how a simultaneous-event multivariate point process (SEMPP) model can be efficiently fit to ensemble neural spiking activity using a multinomial generalized linear model. We illustrate the versatility of the SEMPP model by analyzing neural spiking activity from pairs of simultaneously-recorded rat thalamic neurons stimulated by periodic whisker deflections (Ba et al. Front Comput Neurosci. 2014 Feb 10;8:6). Correctly managing the state of general anesthesia is crucial for providing most surgical and many non-surgical therapies. We show how the state-space point process framework can be used to construct a system for real-time control of the brain in a state of deep general anesthesia termed burst suppression for the purposes of maintaining a therapeutic medical coma Shanechi et al. PLoS Comput Biol. 2013 Oct;9(10)). The hypothalamic-pituitary-adrenal axis, responsible for the regulation of the steroid hormone cortisol, is one of the most important control systems in the body. We use a linear ode model analyzed with a compressive sensing algorithm to accurately recover the pulsatile inputs from the hypothalamus which drive normal variation in the blood levels of cortisol and adrencorticotropin hormone (Faghih et al. PLoS One. 2014 Jan 28;9(1)). Non-stationarity is the rule rather than the exception for neural signals. We discuss a Bayesian spectral decomposition framework—spectrotemporal pursuit—to compute spectral estimates that are smooth in time and sparse in frequency. We illustrate the approach with a spectral analysis of neurons recorded from humans during the transition into anesthesia-induced unconsciousness (Ba et al. PNAS, 2014, Dec 2, E5336). Developing statistical methods that provide accurate dynamic characterizations of neural systems is critical for understanding how the brain and central nervous system represent and transmit information.