Abstracts

Rationale: With recent advances in neuroimaging technology pre-existing brain mapping techniques such as EEG have been revitalized with indications of improved clinical application. Simultaneous EEG and fMRI (SEM) has shown consistent promise, despite technical obstacles produced by the imaging gradient and ballistocardiographic (BCG) noise. Several noise subtraction techniques exist, each based on either filtering frequency bands or waveforms that are fMRI-generated but they do not produce EEG that is consistently or adequately free of noise. Methods: We have employed analog noise reduction to increase the signal to noise ratio (SNR) circumventing the technical barriers presented by post-processing methods by utilizing an electrode cap that gathers two channels of data from each electrode location with a matched reference electrode located at the same scalp site but insulated from it thereby detecting only the ambient noise. The reference electrodes a contact a layer that closely fits the contours of the head matching the conductive properties of the scalp, while remaining electrically insulated from the body. As both imaging gradient and BCG artifacts are produced by magnetic gradients that are anisotropic within the imaging field-of-view, this more spatially selective noise recording allows more accurate subtraction than conventional common mode rejection. SNR is increased further by using all-carbon electrodes and wires, combined with segmented RF filtering enclosures incorporated within the amplifier resulting in a noise reduction by ~60 dB without post-processing or clock synchronization. Remaining gradient artifact is removed using real-time adaptive software algorithms with an advantage over previous generations of this technology having a much smaller SNR to eliminate. Nevertheless, they advance the technology by constructing an adaptive noise template for each channel, updated over short epochs in the frequency domain, resulting 92 dB min. total gradient noise reduction. BCG is removed using adaptive methods to construct, align, and update the BCG templates for each channel, compensating for heart rate and BCG amplitude variation. Results: The total noise reduction compares well with conventional filtering techniques, which commonly result in a 44 - 48 dB noise reduction. The greater than 40 dB decrease in noise resulting from the analog noise reduction produces EEG tracings that are substantially easier to interpret. An EEG figure accompanies this abstract. Conclusions: The quality of fMRI from SEM depends entirely on the EEG accuracy. Misidentification of epileptiform discharges, both false-positives and false-negatives, reduces the likelihood of accurate fMRI depiction of epileptic discharges being compromised by both the modelling of noise as epileptic activity and by the modelling of epileptic activity as background. Each occurs because of the similarity in sharp components that noise and epileptic activity share. As such, the improved EEG recording ability, as indicated by the substantially improved noise reduction, is set to yield more reliable fMRI results.