Is there an upper limit to the number of channels in an EEG system that would provide additional useful data points?

Question

A quick scan of search results seems to indicate that the highest density EEG systems on the market are limited to 256 channels such as Neuvo, SynAmps, and others. Is this an arbitrary limit (either due to system design or practicality of sensor placement) or is it perhaps a case of diminishing returns with any decreased sensor spacing not providing much in the way of additional information in the data points?

If not 256 channels then what would be an upper limit? Am I correct in assuming that the summation of electrical activity that is expressed at the scalp would be spatially quantized due the raised gyri on the brain surface (with any activity in the intervening sulci being masked)? Then again if that were the case would there still be benefit in having the extra spatial resolution running atop the contours of the gyri?

I'm wondering if a system consisting of a full mesh of electrodes would have merit as a project? [this would be for those such as myself that due to either nature or grooming don't have to work around any hair on there scalp]

Answer 1

For EEG, mostly diminishing returns. Because the contacts themselves are relatively far from the neurons generating those signals, there is a lot of correlated signal on adjacent channels. Statistical methods like clustering are often used to treat signals of interest coming from multiple channels as just a single signal. Increasing the spatial sampling density just means more channels will be included in that cluster.

There could be some improvement to SNR, but that would really only be helpful in the case of single bad contacts (effectively you are adding redundancy) and uncorrelated noise. It's also worth considering that more channels means more data to store and analyze which can increase processing time.

There's also the practical issue that equipment for recording EEG can be quite expensive, so if a lab owns a 256-channel system, increasing the density to allow for 512 channels is a substantial investment. If most EEG labs' systems allow for no more than 256 channels to be recorded, there is little incentive to create commercially available caps that record more than 256 channels, which further makes it less likely for someone to upgrade to more than 256 channel recording, etc.

If you move recordings closer to the brain, like in ECoG, or with depth electrodes, much higher sampling densities are relevant, but in that case for human subjects placement will be based on clinical needs rather than some theoretical limit.

As far as the contributions of sulci versus gyri to EEG signals, that's an active field of research and in summary "it's complicated": a lot depends on which neurons in cortex are involved, and how spatial patterns of activity are organized. You can make general statements about, for example, EEG versus MEG, but otherwise you are just making claims about relative contributions. Increasing sampling density on the scalp will have little to no impact on which brain regions contribute most to the signal.