Yesterday’s noise may be tomorrow’s signal: Where are today’s limits for high-resolution high-frequency surface EEG?

Recent intracranial EEG recordings revealed that frequencies above 100 Hz convey signals highly informative for application scenarios as diverse as movement decoding for Brain-Computer Interfaces and focus localization in cases with neocortical epilepsies. While such novel concepts are advancing rapidly, they still rely on technology that has been optimised for the conventional, i.e., low-frequency (< 100 Hz), EEG. However, such clinical routine recordings are compromised by a progressively decreasing signal-to-noise ratio for higher frequencies; in particular, the non-invasive detection of scalp EEG signals at 1 kHz typical for spikes has always been limited by noise contributions from both, amplifier and body/electrode interface. Accordingly, we developed a low-noise recording set-up optimised to map high-frequency EEG and employed it for the analysis of median nerve SEP components which provide an opportunity to monitor non-invasively the timing of population spikes in the human cerebral somatosensory system. 28-channel SEP were recorded using custom-made amplifiers featuring an intrinsically low white-noise level. Furthermore, in order to reduce thermal Johnson noise contributions from the sensor/skin interface, electrode impedances were adjusted to < 1 kOhm. Based on this tailored approach a high-frequency SEP-component with its main energy at about 1 kHz was identified non-invasively which exhibited a distinct spatiotemporal evolution indicating subcortical as well as several cortical generators. Thus, augmenting the standard low-frequency SEP, the surface mapping of 1 kHz EEG signals may represent an essential step towards non-invasive monitoring of multi-unit spike activity.