Добавил:
kiopkiopkiop18@yandex.ru t.me/Prokururor I Вовсе не секретарь, но почту проверяю Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз: Предмет: Файл:
Ординатура / Хирургия / Библиотека им академика М.И. Перельмана / Книга_6027_Библиотеки_им_академика_М_И_Перельмана.pdf
Скачиваний:
0
Добавлен:
31.08.2026
Размер:
29 Мб
Скачать
8 Designing Your EEG Study 95
cognition. Multiple trials of such events are required to obtain a stable response and detect experimental effects. The other kind is the block-design. In this approach, brain activity is tracked over longer periods of time, allowing the study of persistent states of the brain over minutes or hours, for example, vigilance levels or resting state connectivity. Though most studies focus on one of these approaches, event-related and blocked designs are compatible with ea
ch other.
Multiple standard experimental paradigms have been developed to test faculties such as memory, perception, attention, vigilance, or motor control. A repository with freely available scripts for diverse paradigms is the ERP-Core database (Kappenman et al., 2021). Besides standard experimental paradigms, some studies require a
c paradigm to be designed from scratch. Examples include current efforts
speci for experiments to approximate naturalistic stimuli (e.g., video clips), with the aim of increasing the experiments ecological validity. In the same direction, out-of-lab recordings with real-world tasks have become increasingly common, enabled by mobile EEG devices (Cruz-Garza et al., 2017). Some examples are sporting activ­ities
(e.g. Cheron et al., 2016 ) or machinery operation (e.g. Borghini et al., 2014).
These
paradigms should be well justied in the theoretical framework to support
their validity for investigating the neurocognitive processes of interest.
8.4.1.2 EEG Index
The EEG index of choice should reect the psychological or physiological processes
investigation. EEG has the particularity that multiple indexes for analysis can
under be derived from a single measurement. For example, multiple components can be analyzed from an ERP waveform, or multiple frequency bands from spectral anal­ysis, all derived from multiple channels (Cohen, 2014; Luck & Gaspelin, 2017). It is general
ly advised to base the hypothesis on fewer variables that are well supported by the theoretical framework (Luck & Gaspelin, to
the study means adding more variables to the analysis, hence increasing the
2017). Adding more EEG indexes
likelihood of spurious ndings in one of them by chance. When having more variables, more stringent statistical criteria would be required during the analysis phase, which entails the risk of dismissing a relevant observation. Of course, besides the main hypothesis, further explorative analyses are possible, always using suitable statistical techniques (Luck & Gaspelin,
2017; Puce & Hamalainen, 2017).
8.4.1.3 Group/Sample
The sample
from the population must be suitable for obtaining the necessary data. In basic neurocognitive research, participants are most often young, healthy adults, partly because of their accessibility, but also their suitability to research basic cognitive phenomena. To reduce inuence from external variables, participants are screened regarding their health and their cognitive and perceptual status (Keil et al.,
2014). Age, gender, and social status of the participants need to be accounted for,
96 F. Cross Villasana
keeping participants in a comparable age range and with balanced gender represen­tation, while also reporting where and how the participants were recruited (Keil et al.,
2014).
In recent years, numerous calls have been made to increase the diversity of participants in research samples and improve gender, ethnic, and socia l representa­tion (Niso et al., 2022). It is further argued that since participants most often come from
communities near universities and research centers, most samples have W.E.I. R.D. characteristics, standing for Western, Educated, Industrialized, Rich, and Democratic (Nebe et al., 2023; Niso et al., 2022). Such biased samples can limit
generalizability of the observations to the overal l population. Therefore, calls
the have been made for greater efforts to increase diversity in the samples and to reach out to further communities to do research (Niso et al., 2022).
The number of groups and experimental conditions to test is an important factor
to
consider as it affects the kind of statistical tests and the number of trials required to
enhance statistical sensibility (Boudewyn et al., 2018; Puce & Hamalainen, 2017).
different designs can be visualized in Fig. 8.2. A common framework is to test
The two
or more experimental conditions in a single group, called a within-subjects design. When two or more groups are compared in a single condition, it is referred to as a between-subjects design. A combination where more than one group is tested in more than one condition represents a mixed design. In basic neurocognitive research, within-subjects desig ns are generally preferred as they reduce intersubject variabil­ity. In other words, within-subjects designs have greater statistical power. Within­subjects designs can further enhance their statistical power by increasing the number of trials from each participant (Boudewyn et al., mixed
designs also prot from having more trials (or otherwise more data), it has
2018). While between-subjects and
been observed that between-subjects mostly enhances statistical power with a greater number of participants, likely by compensating for intersubject variability (Boudewyn et al., 2018).
In between-subjects and mixed designs, it is important that the overall demo-
c characteristics of groups such as age, education, and gender balance are
graphi comparable. This can be visualized in Fig. match
the participants of the healthy control group with those of the clinical group
8.2. In clinical studies, efforts are made to
(e.g., an educated 30-year-old male patient is paired with an educated male of similar age in the control group). It must also be reported that the groups do not show statistically signicant differences in these variables to ensure they do not act as a confound.
The numbe
r of participants in the groups, together with the strength of the experimental effect, is determinant of statistical power. Power calculators are avail­able to estimate the number of participants required for reliable hypothesis testing according to an expect ed effect size seen in previous reports. If no previous effect size is available, power estimation can be based on a conservative estimate of the expected effect size. Importantly, if EEG is being combined with other techniques or measures, the effect size to consider for power estimation should be based on the measure with the weakest effect (Nebe et al.,
2023).
8 Designing Your EEG Study 97
Fig. 8.2 Experimental designs considering groups and experimental conditions and/or points of measurement

8.4.2 Implementation of the Study

Having identied the strategy and main elements required for testing the hypothesis, the next step is to implement the strategy efciently to obtain data of the highest possible quality. Besides the number of participants, the precision of measurements is an important contributor to statistical power, especially when the ideal sample size cannot be reached for practical reasons (Nebe et al., tuning of the experimental paradigm and of all settings and procedures for data collection. Likewise, it is important to keep in mind the requirements of the planned data processing and statistical analysis, to adjust data collection accordingly (e.g., having a suitable baseline length, collecting enough trials).
2023). This requires the ne-
98 F. Cross Villasana
8.4.2.1 Paradigm/Task Implementation
The chosen experimental paradigm needs to be optimized to the current require­ments.
The aim is to gather sufcient data for analysis while controlling for con­founds. The different parameters that play a role in this are summarized in Table 8.1 and described in more detail below.
In event-related designs, this optimization mainly entails the number of trials. In
iple, more trials improve the EEG signal-to-noise ratio and stabilize random
princ variability. However, too many trials may exhaust participants and alter the brains responses, introducing a confound. Hence, a balance must be found. One must also consider that some trials will be lost due to artifacts or due to participant error. The number of trials per condition is a delicate decision. Useful references for deciding the number of trials are previous studies with similar designs, as well as specic literature about the EEG index being used (e.g., P300 component, theta oscillation, connectivity measures). Some studies have been specically conducted to assess the impact of the number of trials on different EEG signals. For example, Boudewyn et al. (
2018) show complex interactions between the ERP component used, the use of
between-
or within-subjects design, and the effect size to dene adequate trial numbers. But overall, increasing the number of trials improved statistical power. Borras et al. (2022), with various motor signals, suggest 50 trials as a reasonable
ard, while cautioning that adjustments may be needed for different circum-
stand stances (e.g., small effect sizes, clinical populations). Both authors advise having a balance between trial numbers and participant well-being.
The order in which trials from different conditions are presented is also important
to
prevent confounds: If all trials from Condition A are presented rst, and all trials
from Condition B come second, it could happen that the state of the participant
Table 8.1 Trial and block checklist
Trial number/amount of data
Suf
cient to reduce variability and improve signal quality, compensate for loss of trials/data
Not excessive, to avoid overly long sessions
Trial timing
Baseline Trial length: Captures signal of interest Intertrial interval: Enough to recede trial aftereffects and jittered to prevent habituation and
expectation Order of trials
Randomized: Sequential: If justied by the design: E.g., easy condition rst, difcult second
Blocks
Block Number of blocks: Suitable to collect the required trials or sufcient data Counterbalancing: Shufe starting block between-subjects (prevent confounds) Breaks between blocks: Long enough for the participant to recover
length: Sufcient for the signal of interest
To counter learning and exhaustion confounds
length: Appropriate to keep focus
8 Designing Your EEG Study 99
varies between them, affecting task performance. For example, participants might not yet be used to the task during Condition A, while they may be tired for Condition B. Through randomization of the trial order, a balance is achieved where any learning, exhaustion, or repetition effects affect all experimental conditions equally.
The length of each individual trial should be determined by the task requirements and the EEG index of interest. This involves the duration of stimulus presentation, time for the participant to respond, and time for EEG modulations to show. Between trials or within trials with sequences of stimuli (e.g., cue and target), the timing between stimuli needs to be considered. The stimulus-onset-asynchrony (SOA) is the period between the onset points of two stimuli, and the inter-stimulus-interval (ISI) is the period between the offset of one stimulus presentation and the onset of the following stimulus. Both parameters are calibrated to prevent the overlap of neural responses to subsequent stimuli and adjust task difculty (Cohen,
2014; Luck,
2014). The timing between trials is another factor that affects the measurement
and
the length of the experiment. An intertrial interval (ITI) is necessary for any aftereffects to recede, and avoid the overlap of neural processes between trials. Adding a jitter on top of a standard ITI is helpful to prevent habituation and expectation effects. The precise length of these timing parameters depends on the paradigm in question, and it is recommended to base these settings on the literature from the eld in question.
In most cases each trial contains a basel ine period before the stimulus or event, which facilitates quantifying the EEG modulations with respect to a reference period. The length of the baseline varies depending on the EEG signal. In ERP studies, baselines of hundreds of milliseconds are common (e.g. Feuerriegel & Bode,
2022), but the ideal baseline length is still debated (e.g. Delorme, 2023; Feuerriegel
Bode,
& adjus
2022). For time-frequency analyses longer baselines are required, which are
ted according to the wave-length of the oscillation in question, so that baselines
can reach lengths of seconds (Cohen, 2014).
Most neurocognitive experiments are long; therefore, it is necessary to break the stimul
i presentation into separate blocks with equal trial numbers and introduce breaks between blocks for participants to rest and recover. During breaks, partici­pants can stop attending to the stimuli, relax, and move freely. Besides individua l trials, blocks also serve as a way to introduce experimental conditions, for example, in block designs with continuous measurements, but also in event-related designs. Counterbalancing refers to a systematic variation in the way participants are exposed to the experiments settings and conditions most often related to the blocks. For example, in some experimental tasks, the response hand is switched between blocks as a control measure. Counterbalancing in this case implies alternating the initial response hand across participants and switching the response in each succes­sive block. Counterbalancing prevents the response hand itself from becoming a confound. This measure can be used on top of randomization for better variable control. But in certain designs where randomization is not possible, counterbalancing can still be used to prevent confounds. For example, an experiment may require all trials within a block to be Condition A or Condition B without randomizing, but counterbalancing can still be used to alternate which is the initial condition for each participant.
100 F. Cross Villasana
8.4.2.2 Measurement Precision
Recording conditions should be arranged to ensure that noise is reduced to a minim
um, events during the recording are timed with precision, and that the data contains the signal of interest. This requires setting up the hardware to optimize the measurement and reduce artifacts, which is described in Chap. 15 (Getting Clean
Data: Artifacts and How to Prevent Them). However, the personnel who operate the devices are just as important. Operators should be well-trained in setting up the EEG and other devices accurately and in an adequate time frame, so that suitable data are obtained without exhausting participants (Nebe et al.,
tors also need to monitor all devices during acquisition and make adjustments
Opera
2023).
when necessary, for example, check the cap position and electrode impedance during block breaks.
The artifact handling strategy must be considered during the study design,
ing the signal of interest, recording situation, and EEG preprocessing pipeline.
includ To illustrate, one may expect different artifacts from healthy participants in the laboratory, patients in a clinic, or mobile recordings. According to the situation, one may consider reques ting healthy young participants to try to reduce their blinking to avoid this artifact. In contrast, in a clinical study, let patients blink freely to assure their well-being, and later use blink correction techniques on the data. Meanwhile, for mobile or patient data, where it is know n that the data will be noisier, one may opt to record more data/trials compared to standard lab studies, and use advanced data processing techniques.
Some EEG signals require special controls regarding artifacts, and this must be
consi
dered during design. For example, accurate detection of eye movements is required when investigating the gamma oscillation, b ecause activation of ocular muscles generates spikes with energy in the same frequency range (Niso et al.,
2022). Also consider the processing techniques in relation to the signal of interest
recording situation, taking care that the processing applied to the data does not
and drastically affect the signal (Niso et al.,
2022; Pernet et al., 2020).
8.4.2.3 Experimental Protocol
An experimental protocol species the procedures that the operators must follow for collecti
ng data. It indicates the number of required sessions, and within each session, it species the duration and steps to follow (e.g., welcoming participants, signing informed consent, recording data, debrieng).
For counterbala
ncing purposes, the protocol must identify factors such as the experimental conditions that apply to each participant, which block or response mapping is applied rst, or the block order to follow. Requirements for participants must also be specied. This involves control of substance use, such as caffeine, alcohol, or cigarettes (e.g., avoid caffeine 2 h before the experiment, in some laboratories), and accounting for medication status (e.g., on-medication,
8 Designing Your EEG Study 101
off-medication). Other common requirements are having normal or corrected to normal vision/perception, washing hair, avoiding the use of hair products, and having good sleep the night before. The time of the day is another important factor, as circadian rhythms and participantsroutines affect the brain state and perfor­mance. As a general rule, working hours between 09:00 and 17:00 are suitable so participants are not tired or hungry. But optimal timing must be con individual participant, considering their individual routines. Chapter
ow and Lab Management) includes a detailed overview of the nuances of
Work
rmed with each
11 (Study
implementing an EEG experiment and the day-to-day administration of an EEG laboratory.
8.4.2.4 Pilot Testing
Running a pilot test that involves all aspects from participant recruitment to data analys
is is the best way to ensure that everything works as intended before the real experiment. If a problem was detected at any stage, it is possible to correct and adjust settings. You can consult Chap. 10 (Pilot Testing) for a detailed review of the
g process.
pilotin

8.5 Concluding Summary

Careful study design is critical for the success of a research project. It is always worth spending sufcient time on ensuring that all elements are in place to guarantee a successful study. In this chapter, we provided an overview of the most important factors to consider while designing EEG studies. However, it is always advisable to review similar studies to the planned study design to get the most detailed informa­tion on the requirements specic to the area of study.

References

Borghini, G., Astol, L., Vecchiato, G., Mattia, D., & Babiloni, F. (2014). Measuring neurophys-
iological signals in aircraft pilots and car drivers for the assessment of mental workload, fatigue and drowsiness. Neuroscience and Biobehavioral Reviews, 44, 58– 75.
neubiorev.2012.10.003
Borras, M., Romero, S., Alonso, J. F., Bachiller, A., Serna, L. Y., Migliorelli, C., & Mananas, M. A.
(2022).
Inuence of the number of trials on evoked motor cortical activity in EEG recordings.
Journal of Neural Engineering, 19(4), 046050.
Boudewyn, M.
take to get a signicant ERP effect? It depends. Psychophysiology, 55(6), e13049. https://doi.
org/10.1111/psyp.13049
A., Luck, S. J., Farrens, J. L., & Kappenman, E. S. (2018). How many trials does it
https://doi.org/10.1088/1741-2552/ac86f5
https://doi.org/10.1016/j.
102 F. Cross Villasana
Bulajic, A., Stamatovic, M., & Cvetanovic, S. (2012). The importance of dening the hypothesis in
Cheron, G., Petit, G., Cheron,
Cohen, M. (2014). Analyzing neural time series data: Theory and practice. MIT Press. https://doi.
Cruz-Garza, J. G., Brantley, J. A., Nakagome, S., Kontson, K., Megjhani, M., Robleto, D., &
Delorme, A. (2023). EEG is better left alone. Scientic Reports, 13(1), 2372. https://doi.org/10.
Doody, O., & Bailey, M. E. (2016). Setting a research question, aim and objective. Nurse
Fandino, W. (2019). Formulating a good research question: Pearls and pitfalls. Indian Journal of
Farrugia, P., Petrisor, B. A., Farrokhyar, F., & Bhandari, M. (2010). Practical tips for surgical
Feuerriegel, D., & Bode, S. (2022). Bring a map when exploring the
Haynes, B. (2006). Forming research questions. Journal of Clinical Epidemiology, 59(9), 881–886.
Hunt, H., Pollock, A., Campbell, P., Estcourt, L., & Brunton, G. (2018). An introduction to
Jeong, J.-S., & Kwon, Y.-J. (2006). Denition of scientic hypothesis: A generalization or a causal
Kappenman, E. S., Farrens, J. L., Zhang, W., Stewart, A. X., & Luck, S. J. (2021). ERP CORE: An
Keil, A., Debener, S., Gratton, G., Junghofer, M., Kappenman, E. S., Luck, S. J., Luu, P., Miller,
Kivunja, C. (2018). Distinguishing between theory, theoretical framework, and conceptual frame-
Luck, S. J. (2014). An introduction to the event-related potential technique, second edition. MIT
Luck, S. J., & Gaspelin, N. (2017). How to get statistically signicant effects in any ERP
Malik, A., & Amin, H. U. (2017). Designing EEG experiments for studying the brain 1st edition
Nebe, S.,
c research. International Journal of Education Administration and Policy Studies, 4(8),
scienti 170–176.
Zarka, D., Clarinval, A. M., & Dan, B. (2016). Brain oscillations in sport: Toward EEG biomarkers of performance. Frontiers in Psychology, 7, 246.
2016.00246
org/10.7551/mitpress/9609.001.0001
Contreras-Vidal, Evaluation of signal quality and usability. Frontiers in Human Neuroscience, 11, 527.
doi.org/10.3389/fnhum.2017.00527
1038/s41598-023-27528-0
Researcher,
Anaesthesia,
research: 278–281.
multiverse: A commentary on Clayson et al. 2021. NeuroImage, 259, 119443. https://doi.org/
10.1016/j.neuroimage.2022.119443
https://doi.org/10.1016/j.jclinepi.2006.06.006
overviews Systematic Reviews, 7(1), 39. https://doi.org/10.1186/s13643-018-0695-8
explanation? 637–645.
resource for human event-related potential research. NeuroImage, 225, 117465.
open
doi.org/10.1016/j.neuroimage.2020.117465
A., & Yee, C. M. (2014). Committee report: Publication guidelines and recommendations for
G. studies using electroencephalography and magnetoencephalography. Psychophysiology, 51(1), 1–21.
https://doi.org/10.1111/psyp.12147
work:
A systematic review of lessons from the eld. International Journal of Higher Education,
7(6), 14566–14566. https://doi.org/10.5430/ijhe.v7n6p44
Press. https://books.google.de/books?id=SzavAwAAQBAJ
experiment
1111/psyp.12639
design
code and example datasets. Academic.
Reutter, M., Baker, D. H., Bölte, J., Domes, G., Gamer, M., Gärtner, A., Gießing, C.,
Gurr, C., Hilger, K., Jawinski, P., Kulke, L., Lischke, A., Markett, S., Meier, M., Merz, C. J.,
J. L. (2017). Deployment of mobile EEG technology in an art museum setting:
23(4), 19–23. https://doi.org/10.7748/nr.23.4.19.s5
63(8), 611–616. https://doi.org/10.4103/ija.IJA_198_19
Research questions, hypotheses and objectives. Canadian Journal of Surgery, 53(4),
https://www.ncbi.nlm.nih.gov/pubmed/20646403
of reviews: Planning a relevant research question and objective for an overview.
Journal of the Korean Association for Research in Science Education, 26,
(and why you shouldnt). Psychophysiology, 54(1), 146–157. https://doi.org/10.
J., Leroy, A., Cebolla, A., Cevallos, C., Petieau, M., Hoellinger, T.,
https://doi.org/10.3389/fpsyg.
https://
ERP data processing
https://
8 Designing Your EEG Study 103
Popov, T., Puhlmann, L. M. C., Quintana, D. S., et al. (2023). Enhancing precision in human neuroscience. eLife, 12.
https://doi.org/10.7554/eLife.85980
Niso, G., Krol, L. R., Combrisson, E., Dubarry, A. S., Elliott, M. A., Francois, C., Hejja-Brichard,
Y.,
Herbst, S. K., Jerbi, K., Kovic, V., Lehongre, K., Luck, S. J., Mercier, M., Mosher, J. C.,
Pavlov, Y. G., Puce, A., Schettino, A., Schon, D., Sinnott-Armstrong, W., et al. (2022). Good scientic practice in EEG and MEG research: Progress and perspectives. NeuroImage, 257,
119056.
https://doi.org/10.1016/j.neuroimage.2022.119056
Pernet, C. (2015). Null hypothesis signicance testing: A short tutorial. F1000Res, 4, 621. https://
doi.org/10.12688/f1000research.6963.3
Pernet, C., Garrido, M. I., Gramfort, A., Maurits, N., Michel, C. M., Pang, E., Salmelin, R.,
Schoffelen,
J. M., Valdes-Sosa, P. A., & Puce, A. (2020). Issues and recommendations from the OHBM COBIDAS MEEG committee for reproducible EEG and MEG research. Nature Neuroscience, 23(12), 1473–1483.
https://doi.org/10.1038/s41593-020-00709-0
Puce, A., & Hamalainen, M. S. (2017). A review of issues related to data acquisition and analysis in
EEG/MEG
studies. Brain Sciences, 7(6), 58.
https://doi.org/10.3390/brainsci7060058
Ratan, S. K., Anand, T., & Ratan, J. (2019). Formulation of research questionStepwise approach.
Journal
of Indian Association of Pediatric Surgeons, 24(1), 15–20.
https://doi.org/10.4103/
jiaps.JIAPS_76_18
Szucs, D., & Ioannidis, J. P. A. (2017). When null hypothesis signicance testing is unsuitable for
research:
A reassessment. Frontiers in Human Neuroscience, 11, 390.
https://doi.org/10.3389/
fnhum.2017.00390
Thabane, L., Thomas, T., Ye, C., & Paul, J. (2009). Posing the research question: Not so simple.
Canadian
Thompson, W.
Journal of Anaesthesia, 56(1), 71–79.
H., & Skau, S. (2023). On the scope of scientic hypotheses. Royal Society Open Science, 10(8), 230607.
https://doi.org/10.1098/rsos.230607
https://doi.org/10.1007/s12630-008-9007-4
Chapter 9
Applying Statistics in Your EEG Research
Shivakumar Viswanathan
Abstract Electroencephalography (EEG) is used to measure weak electrical poten-
generated on the scalp by the brains activity. However, EEG recordings have
tials several sources of measurement variability due to interindividual differences and the specics of measurement. Therefore, the application of statistical inference has a crucial role in EEG data analysis to detect experimental effects. In this chapter, we provide an introductory overview of the role of statistical infer ence in an EEG study. We discuss the sources of uncertainty in EEG measurements and how EEG analyses are organized to reduce measurement-related error. Finally, we highlight the impor­tance of reducing measurement error, both during acquisition and analysis, and the value of well-dened hypotheses specic to the EEG experiment.
Keywords EEG · Statistics · EEG analys
is · Measurement error · Hypothesis testing

9.1 Introduction

A major goal of experimental research with electroencephalography (EEG) is to evaluate hypotheses about brain function. A key element to achieving this objective involves applying statistics.
This chapter is for readers who are beginning their EEG research journeys and are
faced
with the question: How do I apply statistics in my EEG study? This is an important question as statistics has multiple roles in a typical EEG study. While planning a study, statistical considerations help to turn a scientic question into a structured experiment (see Chap.
data, statistical methods and tests are applied, often extensively, to understand
EEG observed data patterns and their relevance. The inuence of statistics does not end there. When a studys ndings are published in a scientic article, other researchers often use the reported statistical details and their graphical visualizations to judge the
S. Viswanathan () Brain Products GmbH, Gilching, Germany e-mail:
shivakumar.viswanathan@brainproducts.com
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2026 T.
Warbrick (ed.), The EEG Handbook,
https://doi.org/10.1007/978-3-032-20450-9_9
8: Designing Your EEG Study). While analyzing
105