Electroencephalography and functional magnetic resonance imaging are widely used to study brain function, offering complementary strengths in temporal sensitivity and spatial precision. Integrating these methods enables a more comprehensive understanding of neural activity. A crucial requirement for their successful combination is the accurate localization of electrodes, i.e., the precise determination of their spatial positions relative to each individual's anatomical structures. However, this task remains technically demanding and time-consuming. This master's thesis addresses this challenge by developing a new automated processing pipeline that extends the capabilities of the open-source Electrode Localization Kit. The approach introduces standardized three-dimensional head scans, along with advanced computer-vision techniques that employ shape, color, and region analysis, as well as superpixel segmentation. The evaluation of the proposed methodology demonstrates a substantial advance in practical usability. It approximately doubles overall processing speed, reduces the need for manual intervention by a factor of four, and improves reliability through more than a threefold reduction in false electrode detections.
|
