| Description |
This dissertation describes the use of cortical surface potentials, recorded with dense grids of microelectrodes, for brain-computer interfaces (BCIs). The work presented herein is an in-depth treatment of a broad and interdisciplinary topic, covering issues from electronics to electrodes, signals, and applications. Within the scope of this dissertation are several significant contributions. First, this work was the first to demonstrate that speech and arm movements could be decoded from surface local field potentials (LFPs) recorded in human subjects. Using surface LFPs recorded over face-motor cortex and Wernickes area, 150 trials comprising vocalized articulations of ten different words were classified on a trial-by-trial basis with 86% accuracy. Surface LFPs recorded over the hand and arm area of motor cortex were used to decode continuous hand movements, with correlation of 0.54 between the actual and predicted position over 70 seconds of movement. Second, this work is the first to make a detailed comparison of cortical field potentials recorded intracortically with microelectrodes and at the cortical surface with both micro- and macroelectrodes. Whereas coherence in macroelectrocorticography (ECoG) decayed to half its maximum at 5.1 mm separation in high frequencies, spatial constants of micro-ECoG signals were 530-700 ?m-much closer to the 110-160 ?m calculated for intracortical field potentials than to the macro-ECoG. These findings confirm that cortical surface potentials contain millimeter-scale dynamics. Moreover, these fine spatiotemporal features were important for the performance of speech and arm movement decoding. In addition to contributions in the areas of signals and applications, this dissertation includes a full characterization of the microelectrodes as well as collaborative work in which a custom, low-power microcontroller, with features optimized for biomedical implants, was taped out, fabricated in 65 nm CMOS technology, and tested. A new instruction was implemented in this microcontroller which reduced energy consumption when moving large amounts of data into memory by as much as 44%. This dissertation represents a comprehensive investigation of surface LFPs as an interfacing medium between man and machine. The nature of this work, in both the breadth of topics and depth of interdisciplinary effort, demonstrates an important and developing branch of engineering. |
