Electrode performance across hair textures.

The question

In people with characterised hair texture, density or hairstyle, how do scalp-EEG electrode designs and fitting approaches compare with conventional placement for usable recording, fitting time and participant comfort?

The purpose is to identify supported design choices and gaps for future testing. Afro-textured, tightly curled, coily and dense hair are a particular focus, while relevant comparisons across other hair characteristics remain in scope. Race or ethnicity will not be treated as a substitute for measuring hair characteristics.

The intended output is a systematic review, with meta-analysis only for genuinely comparable outcomes. An evidence map will organise the studies; uncertainty or insufficient data will be reported as findings, not filled with assumptions.

Proposed scope

Participants and settings

Human scalp EEG studies in adults, children or infants that characterise hair, report a relevant hair subgroup, or explicitly evaluate an approach for textured hair. Clinical and healthy-volunteer settings will be distinguished.

Interventions and comparators

Electrode housings, contact geometries, caps, clips, adapters and fitting approaches compared with a defined conventional method. Record hair preparation, scalp position, electrode material, amplifier and recording conditions for both groups.

Study designs

Randomised, crossover, paired and other comparative human evaluations. Studies describing hair-related differences without testing an intervention will inform a separate contextual evidence map.

Evidence kept separate

Bench-only work, simulations, commentaries and manufacturer claims cannot supply human performance effect estimates. Include relevant preprints and conference reports in the search, label their status, and link reports from the same participant cohort to avoid double counting.

Cutting, shaving, braiding, chemical treatment and other preparation requirements will be extracted rather than hidden. Sentinel’s design goal is scalp access without cutting or shaving; that preference will not be used to exclude evidence that challenges the proposal.

What we would measure

• Usable recording. Interpretable recording time or retained data, with the original definition, denominator and assessor. Clinician-rated interpretability and algorithmic artefact rejection are separate outcomes.
• Signal and contact quality. Usable channels, noise, signal-to-noise ratio, contact loss and impedance. Each measure stays separate; improved impedance alone is not clinical effectiveness.
• Fitting and maintenance. Setup time with explicit start and finish points, staff involvement, refitting, preparation and interruptions.
• Comfort and harm. Participant-reported comfort, skin injury, pressure, traction, hair damage, acceptability and withdrawals. Unreported harm is not evidence of no harm.
• Clinical relevance. Wear duration, movement, repositioning, montage coverage and clinical setting. Diagnostic performance will be described only where actually evaluated against a stated reference standard.

Usable recording is the proposed primary outcome family. Specific definitions, time points and the hierarchy for selecting multiple reported measures must be fixed in the final protocol before extraction.

Search, selection and appraisal

1. Check overlap and finalise the question. Search existing reviews and protocol registries, including PROSPERO and OSF. Record the added value of this review, or revise its scope if a sufficiently current review already answers the question.
2. Register the agreed protocol. Use PRISMA-P to structure the protocol. Assess eligibility for PROSPERO; OSF Registries offers a general systematic-review registration route. No registration has been submitted.
3. Search reproducibly. Proposed sources are MEDLINE via PubMed, Embase, IEEE Xplore and a citation index such as Scopus or Web of Science, plus relevant preprints, dissertations and reference/citation chasing. Confirm access and finalise database-specific strategies with an information specialist. Save exact queries, search dates and exports. Use no initial publication-date or language restriction, and document any retrieval or translation limitations.
4. Review independently. Two human reviewers independently screen eligibility and extract key outcomes, reconcile differences and use a third reviewer for unresolved disagreements. Record full-text exclusion reasons and resolve duplicate cohorts.
5. Assess bias before interpretation. Use design-appropriate appraisal tools, selected and piloted in the final protocol. Assess recruitment, comparison conditions, ordering effects, confounding, missing data, selective reporting and independence of signal assessment. Do not replace these judgements with a single quality score.

Initial search concepts: EEG/electroencephalography; hair texture, curl, coil, density and protective styles; electrodes, scalp contact, caps and fitting. Named-device searches supplement the broader strategy. This concept list is not an executed or validated database search.

When meta-analysis would be justified

Pool only independent studies answering the same comparison with compatible populations, outcome definitions, measurement units and follow-up. Assess that compatibility before looking for a favourable summary. A random-effects model cannot make fundamentally different questions equivalent.

Infant nets, adult visual-task recordings and prolonged ICU monitoring will not be combined into one performance number. Separate clinical from laboratory settings and account for hardware, hair preparation, montage and wear duration. Preserve paired comparisons; multiple channels, epochs or publications from one cohort are not independent participants.

Where pooling is defensible, the final statistical plan must specify the effect measure, variance handling, model, uncertainty estimates and sensitivity analyses. Missing paired variance will be sought or handled transparently; it will not be invented. If comparability or data are insufficient, use structured tables and a narrative synthesis without a pooled estimate or a count of statistically significant findings.

These decisions follow the principles in the Cochrane Handbook’s meta-analysis guidance. Clinical interpretation and statistical review are still required. The analysis cannot establish equivalence or non-inferiority of Sentinel.

Starting references

These sources helped frame the proposal. They are not a systematically selected study set and have not undergone this review’s eligibility or bias assessment.

1. Lees et al. — Hair type, texture and EEG/ERP data quality. An observational starting point for defining hair characteristics and distinguishing signal measures.
2. Mlandu et al. (2024) — High-density sensor-net design in infants. A comparison of net designs; applicability to adult clinical monitoring needs separate assessment.
3. Kwasa et al. (2024) — Pilot evaluation of Sevo systems for epilepsy. A small comparative pilot with differences in recording conditions that matter for interpretation.
4. Torrens et al. (2026) — Universal EEG Clip. Task-based evidence to assess alongside its population, methods and funding corrigendum.

Ownership, transparency and next steps

Phanes is developing EEG hardware and has a commercial interest in the subject. That interest, any author relationships and any funding must be declared in the final protocol and report. Independent clinical and methodological contributors have not yet been appointed. Authorship will reflect actual contributions.

This draft used AI-assisted literature discovery and drafting. It is not independent dual-reviewer screening or an analysis of extracted study data. Source checks and methods review by the human review team are required before registration.

The proposed outputs are a dated protocol, transparent amendments, a study-selection flow diagram, an extraction table, risk-of-bias judgements and reproducible analysis materials where permitted. We intend to report null, unfavourable and inconclusive findings, alongside any favourable findings. Journal peer review is a future submission step, not a current status.

The next milestone is a clinical and methods review of the question, access to the required sources, and a named review team. Sentinel’s own prototype testing remains a separate programme.