3 ms·

"EEG technology is a crucial diagnostic tool, but the electrical impulses it detects are faint. Commercial headsets like the one I wore were more likely, he sus

"EEG technology is a crucial diagnostic tool, but the electrical impulses it detects are faint. Commercial headsets like the one I wore were more likely, he suspected, to be reading the muscles just below the skin." That is called muscle artifact. In order to get really valuable EEG data, like medical grade, you need many electrodes professionally placed on the scalp. Beyond that, I have seen recordings with subcutaneous (implanted) electrodes recording data on 128 channels at 30khz. That produces a ton of data to parse. The equipment used here are simple toys.

gwern 14y ago
> After his ascent, Mr. Mitra spent some time analyzing data streams from other users. He said it was clear that the headset was, indeed, picking up an EEG signal. But it occasionally seemed to be picking up a different signal, too — one coming from the muscles around the eyes, not the wiring of the brain. As to which signal was driving the ride at that point, without a detailed study it was not possible to say for sure, he said.
jenius 14y ago
Yeah this is a pretty common artifact in EEG studies - it seems like subjects were encouraged to keep their eyes closed for this exhibit, but blink artifacts are often pretty clear. The plus side is that (it least in my experience) when there was a muscle artifact, it was super clear in the data, as the impulses appear much larger than the brain activity and they are easy to filter out, even interactively as the data comes in. You do end up losing that short piece of data though, which is sad. Definitely note that the number of electrodes you use is not a direct correlation to the quality of the data. Using more electrodes will give you better spatial resolution if that's what you're going for. But for many studies and for exhibits like this, spatial resolution makes no difference at all - they are trying to measure overall activity. I have worked with an eeg cap that had 128+ electrodes though - it was cool, took a long time to set up and had lower data quality per electrode. I have produced research grade EEG data which is being published, and subcutaneous electrodes are probably more than you need. I got the data I needed with 10 electrodes, although it did take about a half hour to set up each subject, apply the electrode gel, move hair out of the way, and make sure the data was coming in clean.

gwern 14y ago

> After his ascent, Mr. Mitra spent some time analyzing data streams from other users. He said it was clear that the headset was, indeed, picking up an EEG signal. But it occasionally seemed to be picking up a different signal, too — one coming from the muscles around the eyes, not the wiring of the brain. As to which signal was driving the ride at that point, without a detailed study it was not possible to say for sure, he said.

jenius 14y ago

Yeah this is a pretty common artifact in EEG studies - it seems like subjects were encouraged to keep their eyes closed for this exhibit, but blink artifacts are often pretty clear. The plus side is that (it least in my experience) when there was a muscle artifact, it was super clear in the data, as the impulses appear much larger than the brain activity and they are easy to filter out, even interactively as the data comes in. You do end up losing that short piece of data though, which is sad. Definitely note that the number of electrodes you use is not a direct correlation to the quality of the data. Using more electrodes will give you better spatial resolution if that's what you're going for. But for many studies and for exhibits like this, spatial resolution makes no difference at all - they are trying to measure overall activity. I have worked with an eeg cap that had 128+ electrodes though - it was cool, took a long time to set up and had lower data quality per electrode. I have produced research grade EEG data which is being published, and subcutaneous electrodes are probably more than you need. I got the data I needed with 10 electrodes, although it did take about a half hour to set up each subject, apply the electrode gel, move hair out of the way, and make sure the data was coming in clean.