Abstract
The current lack of diversity in neuroimaging datasets limits the
potential generalisability of research findings. This situation is also
likely to have a downstream impact on our ability to translate
fundamental research into effective interventions and treatments for the
global population. We propose that electroencephalography (EEG) is
viable for delivering truly inclusive and global neuroscience. Over the
past two decades, advances in portability, affordability, and
computational sophistication have created a tool that can readily reach
underrepresented communities and scale across low-resource
contexts—advantages that surpass those of other neuroimaging
modalities. However, skepticism persists within the neuroscience
community regarding the feasibility of realizing EEG’s full potential
for studying the brain on a global scale shortly. We highlight several
challenges impeding progress, including the need to amalgamate
large-scale, harmonized datasets to provide the statistical power and
robust computational frameworks necessary for examining subtle
differences between populations; the advancement of EEG technology to
ensure high-quality data acquisition from all individuals—irrespective
of hair type—and operable by non-specialists; and the importance of
engaging directly with communities to co-create culturally sensitive and
ethically appropriate research methodologies. By tackling these
technical and social challenges and building on initiatives dedicated to
inclusivity and collaboration, we can harness EEG’s potential to deliver
neuroscience genuinely representative of the global population.
Keywords: Electroencephalography (EEG), Inclusive Neuroscience,
Global Neuroimaging, Data Harmonisation, Community Engagement
Much has been written in the social psychological sciences on the
problems that emerge from only sampling from WEIRD populations. The fact
that most of our understanding of brain structure and function comes
from a small segment of the global population is a major cause for
concern in neuroscience and brain health research too (Greene et
al. , 2022). There are clear downstream impacts on translating basic
human neuroimaging research into effective interventions and treatments.
We propose that scalp-recorded EEG, having recentlt celebrated its
centenary (Mushtaq et al. , 2024), is one of the most viable tools
for delivering a genuinely inclusive and global neuroscience.
The last two decades have seen significant advances in the portability
and affordability of EEG, creating new opportunities to reach
under-represented communities and to scale across low-resource contexts.
For instance, brain clocks from diverse populations, modulated by
physical exposome, gender and disease disparities, socioeconomic
inequality, and dementia, can be similar—or even more robustly
assessed—with EEG than with fMRI (Moguilner et al. , 2024). Its
non-invasive nature also makes it suitable for a wide range of
populations. Yet, despite its potential, the neuroscience community
appears skeptical: A recent community survey revealed that many believe
we are decades away from EEG being used as a tool across the globe
(Mushtaq et al. , 2024).
Here, we highlight challenges and suggestions for overcoming the main
barriers. We propose a multifaceted approach focusing on (i) the
amalgamation of large-scale harmonized datasets; (ii) advances in the
technical capabilities of EEG; and (iii) engaging directly with the
communities we aim to study and serve.
Many EEG datasets are already available from different parts of the
world. But such datasets are nothing if not fragmented. Standardization,
e.g., through BIDS, while increasingly common, is far from universal.
Indeed, even when the same tasks are being employed to examine the same
purported constructs, there can be substantial variation in task
parameters. We appreciate the recent efforts in developing benchmark
datasets to provide a helpful starting point for comparison and serve as
a standard for others to follow (Kappenman et al. , 2021).
Extending those beyond individual laboratories and across diverse
settings will help progress the field.
On the signal processing side, advances in data harmonization can help
increase the interoperability and utility of shared data. The HarMNqEEG
algorithm (Li et al. , 2022) developed by members of the Global
Brain Consortium (GBC) provides an example of such an effort. They
introduced a Riemannian approach to harmonizing cross-spectral tensor
data and demonstrated its utility by connecting 1564 datasets from 14
sites collected over 50 years. Advances in computational methods are
also helping to tackle the challenges associated with data diversity-
which can accounted for through data augmentation and deep learning
techniques incorporating perturbations, biological noise, and multimodal
prior information (Ibanez et al. , 2024). The use of synthetic
datasets can also increase representation of groups, enhancing
sensitivity and generalizability (Moguilner et al. , 2024). The
application of generative biophysical models of EEG can account for
heterogeneity and incorporate multiple priors, making the models more
robust to individual differences (Coronel-Oliveros et al. , 2024).
More broadly, probabilistic frameworks including Bayesian and Markov
models enable continuous updating of predictions as new individual data
becomes available (Ibanez et al. , 2024). These frameworks pave
the way for large-scale, harmonized databases that can significantly
enhance statistical and computational power to understand population
differences.
Hardware limitations present another significant barrier. Although EEG
technology has become portable, challenges persist in obtaining
high-quality recordings from individuals with specific hair types
(Bradford et al. , 2024). Standard EEG cap designs often struggle
to maintain good electrode contact, leading to lower signal quality and
frequent participant exclusion. Innovative solutions include comb-shaped
”fingered” electrodes that show promise in improving signal quality by
better adapting to textured hair (see Choy et al. , 2022 for a
review of possible solutions).
While progress in hardware is a critical driver in making EEG viable for
global reach, affordable, high-density systems suitable for widespread
deployment are still lacking. Open-source initiatives and commercially
available low-cost devices are positive, but balancing cost, channel
number and data quality remains a significant challenge. Similarly, the
complexity of operating the equipment, which requires specialized
training, must be reduced. The feasibility of conducting EEG studies
will be limited in low-resource contexts or environments where trained
technicians are not readily available. We need platforms that simplify
setup procedures, automate calibration processes, and provide intuitive
interfaces facilitating data collection with minimal training.
Engaging directly with under represented communities is crucial. Much of
the work in the global north has been collected through convenience
sampling – most highly educated individuals from affluent backgrounds.
We now have the opportunity and privilege to take EEG out to
historically under-represented communities. By involving community
members as partners, we can better understand their priorities and
inform targeted interventions. Culturally sensitive research practices
(e.g., how will you collect data from a Muslim woman wearing a hijab?)
and community engagement strategies (e.g., how will you encourage
participants to take part in studies when they have much more pressing
priorities affecting their daily lives?) will be essential.
Work is already underway to address some of the issues highlighted here.
While the present authors perspective may be biased by the projects we
are personally vested in, efforts such as the GBC, ReDLat, the Global
Brain Health Institute (GBHI), EuroLad EEG, and #EEGManyLabs (Pavlovet al. , 2021) provide examples of how international collaboration
can promote a more inclusive future for neuroscience. The GBC has been
building networks focusing on the unique challenges faced by the global
south. ReDLAT, EuroLad EEG, and GBHI have been illustrating the impact
of macrosocial factors on brain health (Moguilner et al. , 2024).
The #EEGManyLabs initiative is demonstrating the power of multi-site
collaboration, providing a model for achieving the scale necessary to
explore complex questions about brain function. We must build on these
foundations and expand them. This will necessitate investing in
training, building infrastructure, and making multi-site international
collaboration the norm rather than the exception.