Extension of non-invasive EEG into the kHz range for evoked thalamocortical activity by means of very low noise amplifiers.

Ultrafast electroencephalographic signals, having frequencies above 500 Hz, can be observed in somatosensory evoked potential measurements. Usually, these recordings have a poor signal-to-noise ratio (SNR) because weak signals are overlaid by intrinsic noise of much higher amplitude like that generated by biological sources and the amplifier. As an example, recordings at the scalp taken during electrical stimulation of the median nerve show a 600 Hz burst with submicro-volt amplitudes which can be extracted from noise by the use of massive averaging and digital signal processing only. We have investigated this signal by means of a very low noise amplifier made in-house ( minimal voltage noise 2.7 nV Hz(-1/2), FET inputs). We examined how the SNR of the data is altered by the bandwidth and the use of amplifiers with different intrinsic amplifier noise levels of 12 and 4.8 nV Hz(-1/2), respectively. By analyzing different frequency contributions of the signal, we found an extremely weak 1 kHz component superimposed onto the well-known 600 Hz burst. Previously such high-frequency electroencephalogram responses around 1 kHz have only been observed by deep brain electrodes implanted for tremor therapy of Parkinson patients. For the non-invasive measurement of such signals, we recommend that amplifier noise should not exceed 4 nV Hz(-1/2).