Digital recording and review techniques have evolved in parallel with the electronics and data-processing revolution of the past 30 years. Digital recording gained popularity over analog EEGs because of several advantages.^1–3 Digital recording takes advantage of modern microprocessor costs and general flexibility. More specifically, it allows electroencephalogram (EEG)-record review with user-selected montages, filters, vertical scaling (gain or sensitivity), and horizontal scaling (e.g., time resolution or compression). It also replaced warehousing or microfilming paper records and allows for electronic exchange of EEGs. Finally, digital EEG makes possible the routine application of a variety of complex signal-processing tasks, such as frequency analysis, automated seizure detection, statistical quantitative analysis, and dipole source localization.

Recording EEG aims to accurately capture the EEG signals and avoid contaminants. The first step is to boost the microvolt signals recorded at the scalp or occasionally from intracranial electrodes. Amplification by a million-fold or more requires high-quality electronics so as to avoid contamination by other unwanted signals.

Electrode impedances should be kept below 5000 ohms. The amplifier input impedances must be more than 100 megaohms. This allows the amplifiers to work efficiently. Interchannel crosstalk must be <1%, also measured in decibels as 40 dB down or better. Additional electronic noise in the recording should be less than 1.5 μV peak to peak at any frequency from 0.5 to 100.0 Hz, including 50–60 Hz.⁴

EEG amplifiers are differential: they amplify separately (1) the voltage between one recording site and the ground electrode and (2) the voltage between a second recording site and the ground electrode. Then the electronics subtract one from the other (producing the difference between the two). The subtraction process removes other signals (e.g., 60 Hz line noise) that is present and common to both recording electrodes. The subtraction process itself is imperfect, so that some of the contaminants remain. The perfection of removing the contaminants common to both channels is measured by the common mode rejection ratio (CMRR). The higher the CMRR, the better the amplifier system excludes unwanted contaminants.⁵ CMRR is measured in decibels of attenuation; a higher value indicates a cleaner signal. CMRR > 100 dB (attenuation by 100,000) are typical of today’s amplifier systems.

Many systems now record EEG using a single reference electrode as the second recording channel in the differential system. This allows the subsequent software to change montages at will, a process described further below. In such a system, it is critical that the ground and reference electrode remain well connected.

This initial amplified signal is an analog signal, that is, continuous in time and amplitude like most traditional electrical signals. The signal is filtered to remove further unwanted contaminants. The high filter often is set at 100 Hz and a low filter at 0.16 Hz. These reduce some contaminants, such as muscle artifact. It is important to understand that these processes do not eliminate signals outside the desired range; rather, they diminish them. High-amplitude signals, such as muscle artifact >100 Hz, still partially remain.

The amplified, filtered EEG is then digitized. The process involves measuring the signals periodically, one sample per channel at each appointed time. The number of samples per unit of time, which leads to the time between the samples, is the sampling rate. The minimum digital EEG sampling rate is 250 samples/second, but many systems use 400 samples/second or higher.⁴ Some analog-to-digital conversion (ADC) systems digitize one channel at a time, one channel after another. These systems have a measurable time between the first and last channels, which could influence further EEG analysis. Better systems accomplish the ADC very quickly, which avoids this temporal skew between first and last channels sampling.

The ADC sampling rate should be at least twice the fastest frequency present in the analog signal.^4,5 Twice the highest frequency in the amplified, filtered analog signal is referred to as the Nyquist frequency. Any original analog signal higher than the Nyquist frequency will show up after digitization as a distorted signal at incorrect slower frequencies. This is referred to as aliasing. For example, in a system with 250 samples/second, a 550 Hz contaminant will show up aliased at 50 Hz in the digitized signal.

Digitization must use an amplitude resolution of at least 12 bits and must be able to resolve the EEG down to 0.5 μV.^4,5 Larger digital resolution is optional, and many current systems have 16-bit resolution. The higher the resolution, the more accurately the system will capture small details and also be able to record even when the signal varies due to movement of the patient or wires. Higher resolution is therefore an especially useful feature when a patient has a seizure.

The EEG display on a computer monitor should approximate the quality of the visual image on a traditional paper-and-pen EEG system.^1,3,4 The display should approximate the traditional routine EEG time scale that uses 30 mm to represent each second of recording. To do so, each second of EEG displayed should contain at least 100 data points. This can be achieved with a 15-inch monitor with a horizontal pixel representation of 1024 points. Better resolution is preferable, especially for localizing fast waveforms, such as epileptic spikes. Vertical screen spacing between channels should be at least 15 mm. Occasional overlap of data between channels is acceptable. Some epilepsy monitoring units display very large numbers of channels and so must compromise on the vertical resolution preference.

Display systems should display montage designations, gain and filter settings, technologist’s comments, a time stamp, and event markers, along with the EEG data.^1,4 More compressed and expanded horizontal and vertical scales should be available. Side-by-side visual comparison should be available to look at different EEG segments within one recording, as well as different segments from different recordings obtained on different days.

Montages available for review should be consistent with those in standard use and recommended by the International Federation of Clinical Neurophysiology (IFCN) recommendations. For patients with implanted electrodes, arbitrary additional channel labels are needed for those recording points.

Many systems record the EEG in a common reference montage. This allows for digitally reconstructing bipolar of other referential montages.⁴ This is referred to as montage reconstruction or remontaging. It is similar to the analog process previously used in some epilepsy monitoring units,⁶ also referred to as montage reformatting.

Montage reformatting is simply a subtraction of the referential channels from each other. For example, subtracting F3-A1 from Fp1-A1 gives Fp1-F3:

(Fp1-A1) − (F3-A1) = Fp1 − A1 − F3 + A1 = Fp1-F3

Doing so for a large number of channels creates a full bipolar montage even though the originally digitized data were referential. This tactic allows the same EEG data to be reconstructed in many different referential and bipolar montages, which can aid in interpretation of some EEG phenomena.^1,4,6 This is a fundamental advantage of digital EEG over previous paper-and-pen techniques.