Learning to learn: Making sense of electrophysiology data

Abstract

With no previous signal processing background, I began to study how electrical brain waves vary with movement speed and uncertainty. I learned when fleshing out the details or just seeing the big picture made sense for the techniques I used.

My lab uses scalp-electroencephalography (EEG) to record brain activity. EEG data can be noisy, but there are methods see through this notice. After some pre-processing, I ran an independent component analysis to decompose a complex signal into its sub-signals. I removed the eye movement sub-signals as I just was interested in brain activity. With kurtosis values—the sharpness of a signal—I could remove artifactual trials.

I was uncomfortable using ICA and kurtosis measures without knowing exactly how they worked. Learning every nuance would have halted my analysis progression. So, with a conceptual understanding, I used these tools to generate a cleaner EEG signal.

With a clean signal, I began my time-frequency analysis. This would describe how well a sine wave at a given frequency represents my signal. I could not get a conceptual hold on this topic. After pausing my analysis and taking an online course—at my PI’s suggestion—my progress accelerated.

I now could examine how electrical brain activity changes with movement uncertainty and speed. My analysis suggests that brain activity increases with slower movements; however, now I need to learn how to statistically verify this result.