| Description |
Recording the neural activity of human subjects is indispensable for fundamental neuroscience research and clinical applications. Human studies range from examining the neural activity of large regions of the cortex using electroencephalography (EEG) or electrocorticography (ECoG) to single neurons or small populations of neurons using microelectrode arrays. In this dissertation, microscale recordings in the human cortex were analyzed during administration of propofol anesthesia and articulate movements such as speech, finger flexion, and arm reach. Recordings were performed on epilepsy patients who required long-term electrocorticographic monitoring and were implanted with penetrating or surface microelectrode arrays. We used penetrating microelectrode arrays to investigate the effects of propofol anesthesia on action potentials (APs) and local field potentials (LFPs). Increased propofol concentration correlated with decreased high-frequency power in LFP spectra and decreased AP firing rates, as well as the generation of large amplitude spike-like LFP activity; however, the temporal relationship between APs and LFPs remained relatively consistent at all levels of propofol anesthesia. The propofol-induced suppression of neocortical network activity allowed LFPs to be dominated by low-frequency spike-like activity, and correlated with sedation and unconsciousness. As the low-frequency spike-like activity increased, and the AP-LFP relationship became more predictable, firing rate encoding capacity was impaired. This suggests a mechanism for decreased information processing in the neocortex that accounts for propofol-induced unconsciousness. We also demonstrated that speech, finger, and arm movements can be decoded from LFPs recorded with dense grids of microelectrodes placed on the surface of human cerebral cortex for brain computer interface (BCI) applications using LFPs recorded over face-motor area, vocalized articulations of ten different words and silence were classified on a trial-by-trial basis with 82.4% accuracy. Using LFPs recorded over the hand area of motor cortex, three individual finger movements and rest were classified on a trial-by-trial basis with 62% accuracy. LFPs recorded over the arm area of motor cortex were used to continuously decode the arm trajectory with a maximum correlation coefficient of 0.82 in the x-direction and 0.76 in the y-direction. These findings demonstrate that LFPs recorded by micro-ECoG grids from the surface of the cerebral cortex contain sufficient information to provide rapid and intuitive control a BCI communication or motor prosthesis. |
