Seattle, Washington (206)744-3576 or 1-800-374-3267

Use of full band EEG to localize seizures

By John W. Miller, M.D., Ph.D.

The EEG is an irreplaceable but imperfect clinical tool for characterizing seizures and brain function. Standard clinical EEG has limited spatial resolution but has the great advantage of characterizing actual seizures and related phenomena in real time. In epilepsy surgery, long-term video EEG monitoring (LTM) is a key method for localizing the site of seizure origination for possible surgical removal, but frequently, monitoring with surgically implanted electrodes is needed, and there are cases in which even such invasive monitoring does not clearly localize the focus.

This situation has given rise to many attempts to improve EEG technology to increase its detection and localizing power. Dense array EEG is under active investigation at the Regional Epilepsy Center. This method, which combines LTM recording with 256 electrodes and source analysis software, increases the ability of EEG to detect seizures and greatly improves their localization. Another approach is full band EEG, which finds new information in EEG frequency bands beyond the standard 1–70 hertz range used in clinical practice.

The Infraslow EEG

Recording the infraslow EEG (from DC to 1 hertz) requires DC-coupled amplifiers with a large dynamic range, so that they will not be saturated by slow drifts in baseline voltage. Standard EEG electrodes are made of gold, tin, or steel, and have capacitance, which can block lower frequencies. To prevent this problem, silver/silver chloride electrodes, which are reversible and do not polarize, are used. Modern silver/silver chloride electrodes are sintered, with silver chloride particles baked in, so they don’t have to be chloridated and can be autoclaved and reused.

Another issue is distinguishing slow artifacts from the infraslow signals generated by seizures. The galvanic skin response, a standing skin potential in the millivolt range, which is primarily related to sweat gland activation, can be eliminated by puncturing the skin partway with a needle. Artifacts from eye movements and blinking can be identified by their characteristic location. High amplitude, diffuse, very slow potentials can also be created by tongue movements, cough, strain, body tilt, and hyperventilation. The infraslow voltage changes that accompany these physiological maneuvers are too large to be explained by neuronal or glial generators. It has been hypothesized that these slow artifacts are attributable to a large standing voltage difference across the blood brain barrier and result from changes in the size of vascular compartments or in the path of the leakage current from this blood brain barrier potential.

We performed recordings with DC-coupled amplifiers in patients with presumed temporal lobe epilepsy (Vanhatalo et al., 2003). Infraslow signals (30–150 uV) began within a few seconds with every seizure. In seizures with presumed mesial temporal lobe onset (n = 7), the polarity of the slow activity was initially positive but changed to negative after lateral temporal spread of the seizure. In all seven cases, the side of the initial infraslow potential was the same as the lateralization of the seizures determined from other data for surgical planning, including one case in which seizures could not be lateralized with conventional EEG.
Five of the eight patients in this study who received epilepsy surgery saw significant seizure reduction, and infraslow signal analysis correctly localized the region of seizure onset in all five.
These results were confirmed in a subsequent study of 20 temporal and extratemporal seizures in 11 patients, using an expanded 35-electrode array and virtual source montage analysis to assist differentiation from slow artifacts (Miller et al., 2007). Five of the eight patients in this study who received epilepsy surgery had followup examinations that documented significant seizure reduction, and infraslow signal analysis correctly localized the region of seizure onset in all five, while conventional noninvasive EEG recording and analysis localized only three of the five.

We have also detected infraslow shifts in invasive recordings. We analyzed 82 seizures in 11 patients recorded with a 64-channel DC-coupled amplifier coupled to subdural grids with platinum electrodes, at a bandwidth of 0 to 100 hertz. Infraslow signals with amplitudes of 0.8 to 10 mV were seen with seizures in 10 of 11 patients, starting from 2 seconds before to 493 seconds after electrical onset determined by conventional recording. The very high ictal voltage helps explain why infraslow activity can localize seizures on scalp recordings better than conventional EEG.

The Fast EEG

High frequency recording — 70 to 400 hertz and beyond — may be of special value in localizing seizures with invasive recordings. Ripples (80–200 Hz) in CA1 pyramidal cells and entorhinal cortex have been proposed to have a normal functional role, while fast ripples (250–500 Hz) are proposed to be pathological. Fast ripples correlate with sites of spontaneous seizures in rats made epileptic by intrahippocampal kainite injections. Fast ripples are proposed to result from axons of depolarized hippocampal pyramidal cells, which are electrically coupled by gap junctions. Fast ripples are also believed to be due to synchronous action potentials, which generate strong field potentials (“field ripples”) that in turn generate and synchronize action potentials in an autoregenerative fashion.

Ictal fast ripples have been used to localize seizures in human recordings with depth electrodes (Jirsch et al, 2006). Ictal fast ripples were recorded with depth electrodes at onset of partial seizures in seven of ten patients, but not in three patients whose seizures were not localizable from implanted electrodes. Contacts of fast ripples corresponded only to seizure onset zones determined by conventional EEG. This preliminary work raises the possibility that ictal fast ripples may be markers of the seizure onset zone — that is, they may assist in distinguishing the true site of seizure origination from mere regions of seizure spread.


This work indicates that full band EEG has practical value. Most focal seizures have high amplitude focal infraslow shifts that sometimes localize seizures on scalp EEG better than conventional recordings.
There is important information at the extremes of the EEG spectrum, and useful data are discarded when recordings are done at the conventional 1–70 hertz frequency band.
Very fast activity — fast ripples — can sometimes be recorded at the beginnings of seizures, and with more study, fast activity may prove useful in confirming that the genuine site of seizure origination has been found. Full band EEG requires special electrodes and improved amplifiers that are only just beginning to appear in a few commercially available EEG systems. It is now clear that there is important information at the extremes of the EEG spectrum, and that useful data are discarded when recordings are done at the conventional 1–70 hertz frequency band. More investigation is needed to demonstrate conclusively if full band EEG can reduce the need for invasive monitoring or improve surgical outcomes.

Jirsch JD, Urrestarazu E, LeVan P, Olivier A, Dubeau F and Gotman J. High-frequencyoscillations during human focal seizures. Brain. 2006;129:1593–1608.

Miller JW, Kim WS, Holmes MD, Vanhatalo S: Ictal localization by source analysis of infraslow activity in DC-coupled scalp EEG recordings. Neuroimage. 2007;35(2):583-97.

Vanhatalo S, Holmes MD, Tallgren P, Voipio J, Kaila K, Miller JW. Very slow EEG Responses indicate the laterality of temporal lobe seizures: a DC-EEG study. Neurology. 2003;60:1098-1104.