Spatial Brain Control Interface using Optical and Electrophysiological Measures
Abstract
Functional imaging and electrical recordings were performed over posterior parietal cortex during a spatial attention task. Subjects had to covertly shift their attention with simultaneous visual stimulation and motor planning. The imaging response showed modulation for spatial conditions. Various analysis and decoding methods were assessed to extract a prediction signature from this brain activity. The Linear Support Vector Machine (LSVM) was the most appropriate for implementing a reliable brain-computer interface (BCI). The LSVM method was applied to the imaging data with various temporal parameters. All variations proved to be suitable to predict experimental parameters (left vs. right eye movement) from the hemodynamic response over PPC. However, due to the slow hemodynamic signal, performance for the LSVM reached only ~60% during the later task phases. Thus, the subject s response could not be predicted reliably before the actual movement. Electrophysiological recordings (single unit and local field potentials) were performed in the previously imaged regions to allow comparison with the hemodynamic response. These electrical signals especially the local field potentials proved to be fast and strongly tuned for the spatial parameters of the task. Thus, a reliable BCI that can predict upcoming movements or behaviors will need to combine signals from various sources.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 27, 2013
- Accession Number
- ADA595608
Entities
People
- Barbara Heider
Organizations
- Rutgers University–Newark