Stepping stone sampling for retrieving artifact-free electroencephalogram during functional magnetic resonance imaging.

Ballistocardiogram and imaging artifacts cause major interference with simultaneous electroencephalogram (EEG) and functional magnetic resonance imaging (fMRI) recording. In particular, the large amplitude of the imaging artifact precludes easy retrieval of EEG signals during fMRI scanning. Recording with 20,000-Hz digitization rate combined with 3000-Hz low-pass filter revealed the real waveform of the imaging artifact, in which it was elucidated that each artifact peak precisely corresponded to each gradient component and actually had differential waveforms of the original gradient pulses. Based on this finding, to retrieve EEG signal during fMRI acquisition, a blip-type echo planar sequence was modified so that EEG sampling might be performed at every 1000 micros (digitization rate 1000 Hz) exclusively in the period in which the artifact resided around the baseline level. This method, called "stepping stone sampling," substantially attenuated the amplitude of the imaging artifact. The remnant of the artifact was subtracted from the averaged artifact waveform. In human studies, alpha activity was successfully retrieved by inspection, and its attenuation/augmentation was observed during eyes open/closed periods. Fast Fourier transform analysis further revealed that even from DC up to 120 Hz, retrieved EEG data during scanning had very similar power distributions to the data retrieved during no scanning, implying the availability of the high-frequency band of the retrieved EEG signals, including even the gamma band.

Detection limit in low-amplitude EEG measurement.

Electrocerebral inactivity for the determination of cerebral death is defined as no findings of EEG greater than the amplifier's inherent internal noise level when recording at increased sensitivity. A surface biopotential electrode contains two interfaces composed of skin gel (electrolyte) and gel electrode (metal), each forming a noise source. The power spectral density, S(f), of extremely low noise signals was obtained by means of autocorrelation and fast Fourier transformation. Interelectrode resistance, R(f), was measured with synchronous rectification. The formula of equivalent noise resistance R(n) = S(f)/4kT, where k is the Boltzmann constant and T is room temperature in Kelvin, gives a resistance that generates the thermal noise corresponding to the measured S(f). Rn/R is a parameter derived from normalization by R. When Rn/R = 1, measured noise contains thermal noise only. Meanwhile, Rn/R > 1 indicates presence of excess noise, such as 1/f, and tissue noise in addition to the thermal noise. Mean square root (Rn/R) of the scalp noise was 10.8 at 10 Hz, showing existence of excess noise. The study results suggest that it is necessary to take excess noise into consideration in the measurement of low-amplitude EEG for the determination of cerebral death.