Ultra-dense EEG sampling results in two-fold increase of functional brain information
Introduction
The earliest EEG recordings detected electric activity of the whole brain via a pair of sensors positioned across the head (Pravdich-Neminsky, 1913, Berger, 1929). It was soon discovered that EEG varied substantially over the scalp (Adrian, 1934, Adrian, 1935), which instigated simultaneous recordings from several EEG sensors and resulted in the standardized “10–20” electrode placement system of 21 scalp electrodes (electrode separation > 6 cm) adopted half a century ago (Jasper, 1958). Although this system remains in wide use among clinicians, high-density (64–256 sensors, electrode separation > 2 cm) EEG systems, which appeared two decades ago (Gevins, 1990, Tucker, 1993) quickly became a popular choice for EEG scientists and have since made their way into some clinics. Given the advances in amplifier miniaturization and wireless data transmission, EEG systems with up to 1000 sensors are well within reach for modern technology. The increased spatial information could significantly impact basic neuroscience research as well as a broad spectrum of practical applications from brain–computer interfaces to localization of epileptic sources. Hence, the important question: how many sensors are sufficient to capture all information on brain activity provided by EEG?
Theoretical and experimental analyses have shown that at least 128 EEG scalp channels are necessary to meet the Nyquist requirement for spatial sampling (Srinivasan, 1998, Spitzer, 1989), but the upper limit remains unknown due to our incomplete knowledge of the electric properties of head tissues and difficulties in accurately modeling real heads. Computational modeling using simplistic head models (Ryynänen, 2004, Ryynänen, 2006, Malmivuo, 2011) and some preeminent EEG methods books (Srinivasan, 2005, Nunez, 2006) suggest that electrodes spaced by 2–3 cm, characteristic of hd-EEG sensor arrays, can provide sufficiently dense EEG sampling, implying that further resolution is unnecessary. A more realistic head modeling based on the finite elements method (FEM) applied to MRI images indicated that a much smaller (~ 0.3 cm) electrode spacing is needed for extraction of cortical patterns from scalp EEGs in humans (Ramon, 2009). This modeling result agrees with experimental measurements demonstrating that electrode spacing of 0.5–1 cm is required to capture spatial EEG pattern without undersampling (Freeman, 2003, Odabaee, 2013). Although these experiments demonstrated that human EEG contains significant high spatial frequencies, it remains unknown how much of the extra information captured at these frequencies is of practical significance, e.g., whether it provides more accurate EEG source localization or improves discrimination between brain states. Here, we use a classification paradigm to show that sampling EEG signals at high spatial frequencies significantly increases classification accuracy derived from EEG responses evoked by two visual stimuli. This demonstrates that ultra-dense (1 cm or less) electrode spacing is needed to fully capture functional brain information from EEG.
Section snippets
Paradigm
Arguably, the best way to determine the sufficient number of EEG sensors is to measure the amount of functional brain information I contained in EEG data as a function of the number of sensors, I(n). For some value of n the function might saturate sufficiently to deem further increase of the number of sensors unnecessary. However, covering the whole scalp with hundreds of sensors is a challenging technical task. Similarly to previous studies (Freeman, 2003, Odabaee, 2013) the compromise we
Results
We used a prototype ud-EEG array: a square 4 × 4 grid of small diameter EEG electrodes with inter-electrode separation of 1 cm (Fig. 1a, Section 2.4). A signal classification paradigm was used to demonstrate that the ud-EEG array provides additional information about brain activity compared to hd-EEG. In this paradigm, eleven subjects viewed images of words presented one by one on a computer monitor (Stimuli section). On each trial, a word appeared on the screen for half a second followed by a
Discussion
The importance of a higher spatial resolution is becoming increasingly appreciated. Our results are consistent with a recent study, where a linear ultra-dense array (2.5 mm inter-electrode spacing) was used to demonstrate that sensor distances 6–10 mm are required to capture the full spatial texture of EEG signal on neonatal scalp (Odabaee, 2013). While one could argue that large fontanels over which the linear array was positioned could cause the increased spatial variation of EEG for neonatals,
Summary
To summarize, we observed surprisingly strong variations of VEPs at 1 cm scale. This result explains a nearly two-fold increase in the classification SNR between the most commonly used hd-EEG sensor densities (~ 0.1 cm− 2, see the Comparison with existing hd-EEG systems section) and the density of sensors in our prototype ud-EEG array (1 cm− 2). Given the advances in amplifier miniaturization and wireless data transmission, ud-EEG helmets with 700–800 sensors needed for full scalp coverage at 1 cm− 2
Acknowledgments
This work was supported by NSF-IIP-1264216, NSF-DGE-0965843 and a Northeastern University Tier1 award grants.
Conflict of interest
The authors declare that they have no competing financial interests.
References (25)
Evaluation of the distortion of EEG signals caused by a hole in the skull mimicking the fontanel in the skull of human neonates
Clin. Neurophysiol.
(2005)Spatial spectra of scalp EEG and EMG from awake humans
Clin. Neurophysiol.
(2003)Mechanical properties of cranial bone
J. Biomech.
(1970)Spatial patterning of the neonatal EEG suggests a need for a high number of electrodes
Neuroimage
(2013)Effects of local skull inhomogeneities on EEG source estimation
Med. Eng. Phys.
(1999)Spatial sampling of head electrical fields: the geodesic sensor net
Electroencephalogr. Clin. Neurophysiol.
(1993)The Berger rhythm: potential changes from the occipital lobes in man
Brain
(1934)The origin of the Berger rhythm
Brain
(1935)Conductivities of three-layer human skull
Brain Topogr.
(2000)Conductivities of three-layer live human skull
Brain Topogr.
(2002)
Über das Elektrenkephalogramm des Menschen
Arch. Psychiatr. Nervenkr.
Beyond topographic mapping: towards functional–anatomical imaging with 124-channel EEGs and 3-D MRIs
Brain Topogr.
Cited by (28)
High-resolution EEG
2019, Handbook of Clinical NeurologyCitation Excerpt :Zelmann et al. (2014) also noted that another reason for the detection of HFOs despite their low amplitudes is that their frequencies are beyond the frequencies that generate the noise in the signal and thus have favorable signal-to-noise ratios. Recently, Petrov et al. (2014) argued for an even higher number of electrodes. They demonstrated in an event-related potential study that a spatial sampling of the EEG with a 1-cm interelectrode distance (corresponding to around 760 electrodes) revealed almost twice the amount of functional brain signals as compared to sampling at 3-cm scale (corresponding to approximately 100 electrodes).
EEG source imaging of epileptic activity at seizure onset
2018, Epilepsy ResearchCitation Excerpt :The 256-channel HD-EEG may not achieve an optimal spatial sampling of the cortical activity at seizure onset (particularly for focal features such as HFOs). Both classical physical modeling of human head tissues (Malmivuo and Suihko, 2004; Ryynanen et al., 2004) and measurement of the human EEG with very closely (3 mm) spaced electrodes (Freeman et al., 2003; Odabaee et al., 2013) have suggested that the optimal inter-electrode distance for the surface of the adult head is less than 1 cm, requiring 500 or more channels (Petrov et al., 2014). Although simulation studies show that the improvement of localization accuracy with increasing channel density is asymptotic (Chu, 2015; Song et al., 2015), with decreasing returns for very high channel densities, it is also the case that important noise characteristics (including both electrode noise and electronic noise) vary independently across channels.
Biomechatronic applications of brain-computer interfaces
2018, Handbook of BiomechatronicsA comparative study of electrical potential sensors and Ag/AgCl electrodes for characterising spontaneous and event related electroencephalagram signals
2015, Journal of Neuroscience MethodsCitation Excerpt :However, Odabaee et al. (2013) found that sensor distances of between 5 and 10 mm are required to capture the full spatial texture of the raw EEG signal on a neonatal scalp. Petrov et al. (2014) found that even when using ud-EEG with a relatively small inter-electrode spacing of 1 cm, there were strong variations in the EEG signal, measured using VEPs. Further, the use of this array led to a two-fold increase in the signal to noise ratio compared to a standard hd-EEG system.