When choosing an electroencephalogram (EEG) device it is important to find the best possible performance by balancing device features and value. Once all the signal related features (those which apply to both sensor and amplifier stages) have been addressed, we will discuss other important characteristics such as connectivity, power, autonomy, data format, etc.
Overview of EEG technical features
Electroencephalography (EEG) technical features can be divided into three main areas:
- The sensor or headset area (sensor layer).
- The amplifier (acquisition layer).
- The connectivity and data layers (connectivity layer) and other important features such as power supply, auxiliary I/O signals, internal sensors, etc.
In this post, we will focus on the third area.
EEG device features
1. EEG power supply and autonomy
The way that the device is powered determines many of the features related to the experimenter's usability, participant mobility, recording duration, etc. There are two possibilities:
- Cable powered: The amplifier is connected directly to the power line through a power supply. The main advantage is that the experiment can last as long as necessary, with no limit on functioning time imposed by a battery charge. The main drawback is that the amplifier must be plugged into an outlet, so the participant will be tethered with a power cable that will constrain both local movement and reach. In the case of walking scenarios, the power cable must be carefully managed to avoid risks of falls or equipment damage.
- Battery powered: The amplifier is powered by batteries, so it can be attached to the participant, allowing free movement and more natural behaviour. The main drawback is that the maximum recording duration will be limited by battery capacity. With a battery-powered amplifier, the battery can be internal or swappable.
- Internal battery: This option maximizes the battery capacity while reducing the required space, and thus the amplifier size. This is because the complexity of the battery housing is minimized. The main disadvantage is that, once the battery is discharged, the amplifier will have to be turned off and recharged, usually for hours, before data collection can be resumed.
- Swappable battery: This second option needs more space for the docking and charging mechanisms, so the amplifier is larger in size. The main advantage is that, when the battery is fully discharged, it can be replaced by a charged one and the recording can be continued immediately. Regarding the battery change, it can be done in two alternate ways: 1) the device must be turned off before the battery is changed and the recording resumed; or 2) the battery can be changed without turning off the device (hot swappable).
The choice between cable-powered or battery-powered amplifiers is usually mediated by the degree of mobility and ecology required for the final application, as batteries are much more adapted to enable freedom of movement. In the case of internal batteries, it is desirable to have 8+ hours of continuous recording time in order to cover a whole day of experiments (or a whole night of sleep experiments). Regarding swappable batteries, they are the best option in recording scenarios where there is no time or means to charge the batteries, or with a very long recording duration.
2. Wired vs wireless EEG amplifier/computer connection
Like the power supply, the amplifier/computer connection has an impact on the freedom of movements for the participant. There are two types of connection, wired or wireless.
- Wired: Most wired EEG devices, if not all, use USB technology. The main advantages of this kind of data transfer is that there is almost no limit in the amount of EEG data that can be transmitted, in real-time, to the computer - with very small latency and high robustness against packet loss or corruption -. The main drawback is the same as with any wired interface, there is always a cable that constrains mobility.
- Wireless: Modern wireless communication technologies convert the amplifier into a mobile device that can be attached to the participant. The two main standards are Bluetooth and WIFI, plus there are manufacturers who use their proprietary technology.
- WIFI: This technology has high throughput (amount of data transferred per second) but at the cost of high power consumption. This requires the use of higher capacity batteries than those devices that use other communication technologies, which then usually leads to larger and heavier devices. The specific communication protocol implemented by the device in WIFI networks is also important. There are two possibilities: TCP or UDP. The advantage of TCP with respect to UDP is that it has mechanisms to guarantee that EEG data is transported to the computer reliably, avoiding data loss, duplications or misalignment in data order. In addition to this, both TCP and UDP protocols have high delays between packets due to the low level infrastructure of the network (TCP more than UDP due to internal data checking). The use of WIFI technology in EEG recording is less common today and not ideal for real-time EEG signal analysis.
- Bluetooth: This has lower throughput than WIFI, but also lower latencies, which makes it a suitable alternative for real-time streaming devices. The communication protocol also requires less power, so smaller batteries can be used, minimizing the amplifier size and weight. The bluetooth standard can be subdivided into two categories: 1) Classic bluetooth, which is designed for continuous streaming data applications; and 2) Bluetooth low energy (BLE), which is designed for low data rate and low power applications that can run on a small battery for long periods of time (note that new versions from 4.2 onwards are increasing data throughput at the cost of increasing power consumption).
- Proprietary technologies: These are technologies developed for ad hoc communications by the manufacturer. This kind of technology uses a dedicated and optimized protocol that usually reduces data transfer (for instance eliminating data headers required by standard communications) in order to maximize data throughput, and sometimes reducing power consumption. The drawback of this kind of technology is that it requires the use of a special dongle or other hardware device to convert the proprietary communication protocol to a universal one, adding complexity to the set up and additional parts that can be broken.
Note that all wireless EEG headsets are prone to disconnections between the emitter and receiver when there is coverage loss or if the electromagnetic space is very “crowded” (i.e. many devices like cell phones or computers, using the same type of connections at the same time in the same place). Thus it is very important that EEG systems implement specific protocols that guarantee immediate reconnection and preservation of data integrity in the event of a connection loss.
If mobility is not necessary, then wired connections can be the choice because of their very low packet delay and high data throughput. However, when mobility is needed, wireless EEG headsets must stream data continuously and without high delay between packets, thus WIFI is only suitable for very specific applications. In the rest of the situations classic bluetooth technologies are the best adapted because they do not need external adapters, do not have a large influence in amplifier size and weight (small and lightweight amplifiers can be designed with 8+ hours autonomy) and can transmit the data without high packet delays. Also nowadays and thanks to the latest improvements in bluetooth low energy protocols commented before, this technology has promising features to become also a standard in EEG devices for real-time data streaming.
3. EEG Amplifier auxiliary I/O signals
Researchers often synchronize EEG electrical signals with external events or stimulation, other biosignals/behavioural measurements, or other devices and applications. Although we can always synchronize by software, the most accurate way to do it is by hardware. This is the reason why most EEG systems have additional external auxiliary inputs/outputs:
- Analog input:
- Visual triggers/photodiodes that measure changes in luminosity, useful to synchronize EEG with visual stimuli on the computer screen, or with strobe lights.
- Auditory triggers/microphones to synchronize with auditory stimuli.
- Analog signals like ExG (electrocardiogram, electrooculogram or electromyogram), galvanic skin response, or other biosensors like respiration effort bands, air flow sensors, etc. which are useful for synchronizing EEG with other physiological responses.
- Digital I/O:
- Pushbuttons or pedals that can detect participant actions, for example in self-paced or cue-based movement experiments.
- Computer I/O (like parallel port) to amplifier adapter to synchronize events coming from computers or other hardware.
Amplifier photo above: 32 channels EEG frontal view. Inputs from left to right: 1x photodiode, 2x ExG, 1x 3-bit digital input (upper row); EEG (lower row)
Auxiliary inputs are an important add-on to any EEG device, providing hardware sample-level synchronization for all recorded signals. These features are very important in wireless amplifiers, as these produce higher delays and time uncertainty on data transmission, which adds to the delay/uncertainty of the receiver software layer.
4. Internal sensors in EEG amplifiers
Modern ambulatory and mobile EEG headsets include movement sensors in the amplifier to facilitate the detection or filtering of motion artifacts in the posterior data analysis steps. The two main sensors used are.
- 3-axis accelerometer: This sensor measures acceleration or G-force in each of the three space axes.
- 6 or 9-axis inertial measurement unit (IMU): 6-axis IMUs have a 3-axis accelerometer and a 3-axis gyroscope that measure rotational motion in each of the three axes. 9-axis IMUs have those previous sensors plus a 3-axis magnetometer that measures magnetic field intensity in the three axes (usually the earth’s magnetic field). There are two advantages of this kind of sensor over the accelerometer: 1) it provides the orientation of the body part moving with the amplifier; and 2) it provides 6 additional degrees of freedom for filtering techniques, increasing the channel space for widely used filtering techniques such as Independent component analysis (see dedicated EEG artifacts post).
For stationary EEG devices, integrating these sensors into the amplifier is not as useful. However, when the amplifier is mobile they are very helpful as a way of improving data analysis, as they give an accurate estimation of the participant’s movements. Inertial measurement units (IMU) are always preferred over accelerometers.
5. Data backup in EEG amplifiers
There are three relevant recording situations where it is useful to record and store the EEG signals in the amplifier itself (data backup), to download it later:
- When sampling at a high rate (large amount of data) but the amplifier connection with the computer has a limited throughput rate.
- When recording with an unstable connection that can potentially be interrupted causing data loss (for instance in scenarios where we share the connection with other devices).
- When the computer cannot be close to the amplifier, which needs to operate like a holter (only recording function for long-term or ambulatory monitoring).
This backup can be implemented in two ways.
- Internal storage: The data is saved in the amplifier internally. The amplifier must be connected to the computer later to download the data. The main drawback of this option is that the amount of internal memory is fixed and can not be increased over the next years of use or swapped out in the field.
- Removable storage (file system mode): The data is saved in a removable storage that can be read from any computer. If more memory is needed, a larger card can be used.
In summary, data backup is usually not necessary when a wired connection is used, but in wireless technologies it provides the experimenter the peace of mind that the data will still be fully recorded if there is a connection problem. In these cases, a removable storage option is an easier and more flexible way to transport the data, which makes it more useful.
6. Output EEG datafiles
EEG data needs to be stored, usually together with some metadata information such as the anonymized participant identifying information or events of interest, for analysis and interpretation. EEG data file formats can be broadly divided into either proprietary or standardized formats. The need for data standardization is driven by the benefits of interoperability: the data sharing in healthcare information systems, and the analysis across research laboratories using software analysis tools (see for instance post on artifact filtering tools). This is of special interest when dealing with large EEG datasets (see post on EEG AI techniques and datasets).
- Proprietary formats. Many hardware manufacturers store the EEG data using their own file format. This provides the manufacturers with a high degree of flexibility at a cost of interoperability with third-party software tools. A strategy to partially overcome the lack of interoperability is to provide software plugins for each of the different software applications to load specific proprietary EEG data files.
- Standardized formats. The two main standard file formats are EDF (Kemp, 2003) and GDF (Schlögl, 2006). These formats allow multichannel data recordings and different sampling rates per signal, and contain a header to store metadata information relative to the participant, amplifier settings, events of interest, etc. They are supported by the main software analysis tools: EEGLAB, FieldTrip, and MNE (see our previous post for a description of them).
Here are examples of real-world EEG montages and artifact tests of two commercial systems, with their respective features (EEG monitoring). The main value with respect to other EEG technologies is practicality for the researcher and comfort for the final user, without having to sacrifice signal quality.
- Diadem EEG: designed to be the first reliable, high signal quality EEG system that can be set up on oneself. The Diadem is used for neuroscience and other applications that require data from frontal and posterior brain areas.
- Versatile EEG: designed to be the most practical EEG system for research, providing excellent signal quality even in the presence of environmental or participant artifacts.
About the author
Aitor Ortiz, MEng. - Research Engineer at Bitbrain (Linkedin)
Aitor Ortiz obtained his degree in telecommunication engineering (2012) and a MSc in biomedical engineering (2013) by the University of Zaragoza (Spain). From 2012 to the present he works as a research scientist and the chief of the electronics department for Bitbrain. His main research interests are on the development of bio-potential recording devices, mainly focused on electroencephalograms for brain-computer interfaces.
- Kemp, B., & Olivan, J. (2003). European data format ‘plus’(EDF+), an EDF alike standard format for the exchange of physiological data. Clinical Neurophysiology, 114(9), 1755-1761.
- Schlögl, A. (2006). GDF-a general data format for biosignals. arXiv preprint cs/0608052.
You might also be interested in:
- What is EEG and what is it used for?
- EEG Electrode Placement: Fixed vs. Variable
- What is QEEG brain mapping and normative databases?
- How to select a dry-EEG headset for your research application
- The Wet EEG Cap & Differences Between Water-Based, Saline and Gel EEG caps
- An introduction to brain-computer interface using EEG signals
- How Deep Learning is changing machine learning AI in EEG data processing
- Bringing BCIs to the user’s home for neurorehabilitation and assistive applications
- Modern BCI-based Neurofeedback for Cognitive Enhancement
- Epilepsy and EEG seizure-detection
- The Forgotten History of Alpha Brain Waves
- The Use of EEG for ADHD Diagnosis and Treatment
- The Procedure and Uses of the EEG Test