Science
200+ peer-reviewed publications. Independent validation against clinical PSG using a 41-gram headband comfortable enough for multi-night home studies. Closed-loop sleep stimulation validated in interventional protocols. The world's largest EEG normative database. A foundation brain model trained on 16.8 million sessions.
Sleep staging validation
56 adults underwent simultaneous Muse S + Level 1 polysomnography. Muse data scored by automated algorithm; PSG scored by blinded registered technologist. Validated across healthy sleepers and participants with sleep-related breathing disorders.
substantial to near-perfect agreement across all sleep stages.
Published in SLEEP Advances, Oxford Academic, 2025. Canadian Sleep Research Consortium. University of Ottawa Institute of Mental Health Research.
N1 agreement is fair — consistent with known difficulty of N1 scoring even among human experts (inter-rater N1 agreement is typically 0.30–0.50). All other stages achieve substantial to near-perfect agreement.
Jaoude et al. Deep learning models (DSN and MNet) achieved 85–86% accuracy and Kappa ~0.78 across Wake, N1–N3, and REM. Trained on 200 recordings, validated on 41. AASM-aligned.
Krigolson et al. Validated Muse for ERP research: N200, P300, and reward positivity quantified. Setup under 10 minutes vs. 35 minutes for $75,000 clinical-grade systems.
Largest at-home EEG study to date. Demonstrated meaningful brain aging biomarkers captured from consumer-grade EEG across diverse populations in unsupervised home environments.
Muse S EEG sleep support showed 20% improvement in Pittsburgh Sleep Quality Index vs. controls. Up to 55% faster time-to-sleep vs. baseline in first 3 nights.
Validated applications
Publications
Curated by research domain. Sleep studies first.
Compared machine-learning and deep-learning methods for EEG-based evoked emotion classification.
Used evolutionary methods to optimize EEG classifiers for brain-machine interaction tasks.
Pilot study using wearable EEG to support reading engagement analytics and personalized learning feedback.
Applied deep-learning methods to EEG brainwave data for emotional sentiment analysis.
Randomized waitlist-controlled trial of EEG neurofeedback for chronic PTSD, reporting PTSD symptom improvement and better affect regulation versus waitlist control.
Presented a scalable system for synchronized EEG recording from audience members in real-world settings, supporting group-level EEG research outside the lab.
Evaluated a short virtual-reality mindfulness meditation training session for sustained-attention outcomes measured with behavioral and EEG-related signals.
Recorded resting-state EEG using a Muse headband and classified trait anxiety into two- and three-class groupings using machine-learning models.
Preprint evaluating Muse portable EEG for rapid stroke assessment and severity prediction.
Developed and evaluated automated sleep staging from wearable Muse EEG, supporting scalable sleep-stage analysis.
Used large-scale Muse EEG recordings from sleep and meditation sessions to predict age and explore mobile EEG biomarkers.
Introduced a method for estimating enjoyment from EEG data collected using a commercial headband.
Large-scale mobile EEG study characterizing age-related differences in EEG dynamics across adulthood.
Studied whether pre-performance EEG beta power is associated with baseball batting performance.
Validated Muse as a low-cost portable EEG system for ERP research, including N200 and P300 measurement in a visual oddball paradigm.
Used Muse EEG data and machine-learning models to classify cognitive fatigue from selected EEG features.
Discussed emotional-behavior mechanisms and applications of wearable devices to mindfulness and emotion regulation.
Used repeated mobile EEG recordings to study individual alpha frequency variability and its relationship with self-reported anxiety.
Systematic review of deep-learning approaches for EEG analysis across applications including sleep, BCI, and clinical monitoring.
Investigated theta-band EEG fluctuations during facial/emotional short-term-memory tasks as potential dementia-related digital biomarkers.
Used recurrent neural networks to classify imagined digits from EEG data for brain-computer-interface applications.
Applied a stacking ensemble approach to classify emotional states using EEG features.
Used EEG during laparoscopic surgical simulation to compare concentration and stress-related brain activity between participants with different experience levels.
Pilot study using portable EEG during meditation exercises in a smart-home setting for people across the cognitive-impairment spectrum.
Evaluated a consumer-grade wearable EEG headband as a feasible tool for assessing state and trait mindfulness.
Assessed user adaptation over time in a brain-computer-interface game environment.
Compared neural and cognitive effects of exercise performed indoors versus outdoors, suggesting environmental context can influence brain-related outcomes.
This study evaluated the use of wearable EEG devices to monitor cognitive load in clinical settings. Findings indicated that these sensors could reliably detect variations in cognitive load during different clinical tasks, suggesting their potential for real-time monitoring in healthcare environments.
Examined how high-intensity interval exercise affects reinforcement-learning behavior and related neuroelectric measures.
Used multimodal physiological signals, including EEG, during public-speaking stress tasks to classify stress responses and explore real-time stress detection.
Used EEG responses to music tracks to classify stress states, showing that EEG features can reflect stress-related responses to auditory stimuli.
Open-label trial evaluating daily Muse-S use for stress reduction among patients experiencing Long COVID symptoms, with feasibility and satisfaction outcomes.
Preliminary study examining frontal brain activity patterns in relation to planned versus impulsive suicide attempts.
Used Muse 2 EEG data and machine-learning models to classify stroke-related EEG signals.
Evaluated a 90-day Muse-S mindfulness program for healthcare professionals during the pandemic, reporting improvements in stress, resilience, cognitive performance, and quality-of-life measures.
Combined handwriting dynamics and quantitative EEG features to classify mild cognitive impairment.
Mixed-methods pilot study examining mindfulness meditation in an interprofessional CPR training context.
Assessed mobile quantitative EEG in children before surgery and found frontal EEG alpha measures were feasible, reliable, and associated with preoperative anxiety measures.
Studied how guided visual-art viewing modulates neural activity measured with EEG.
Presented a neuroadaptive AI chatbot that uses Muse EEG engagement signals to customize learning experiences.
Used segmented EEG recordings from a Muse headband and machine-learning feature sets to classify perceived stress levels.
Used Muse EEG while participants played mobile games to classify engagement states such as bored, engaged, and excited.
Large-scale study using Muse EEG during a short visual oddball task to measure ERP/EEG features and their relationship with perceived cognitive fatigue.
Used wearable monitoring to evaluate how sensory stimulation can modulate cognitive states and task performance.
Introduced a dynamic spatial-filtering approach to improve EEG model robustness when signals are noisy or corrupted.
Introduced a cross-platform tool for designing and running online EEG experiments.
Combined virtual-reality speech interaction with wearable-device measures to explore screening for mild cognitive impairment.
Compared Muse-S automated sleep staging with simultaneous level 1 polysomnography in adults; reported substantial full-night agreement and stage-level accuracy across sleep stages.
Compared Muse measurements with research-grade EEG for spectral analysis and frontal alpha asymmetry, supporting use in selected EEG research contexts.
Used EEG recorded with a commercially available Muse headband to classify perceived stress states. The study supports wearable EEG as a viable signal source for stress detection.
Used EEG recorded during stress-inducing and non-stress-inducing video clips to compare MLP and LSTM models for stress classification.
Used EEG during naturalistic face-to-face interactions to study links between trait mindfulness and inter-brain coupling.
Compared consumer wearable technologies, including Muse, with laboratory systems for assessing workload, stress, and emotional state during work-like tasks.
Applied self-supervised learning to clinical EEG data to learn signal representations from large-scale recordings.
Systematic review of neurofeedback for PTSD; most included studies reported improvement on at least one behavioral outcome, while limitations included small samples and varied methods.
Assessed whether a short Muse EEG recording could help predict stroke severity and support rapid stroke assessment.
Pilot study testing whether portable EEG can differentiate acute ischemic stroke patients from healthy controls using brain-symmetry measures.
Randomized controlled trial evaluating Muse-guided meditation for women newly diagnosed with breast cancer, with outcomes related to stress, fatigue, and quality of life.
Studied technology-supported mindfulness for OCD, including self-reported mindfulness and EEG correlates of mind wandering.
Evaluated brain-sensing devices as support for mindfulness practice, including cognitive and electrophysiological evidence related to attention and mindfulness.
No publications in this category yet.
Foundational brain model
Built over 15 years and trained on 16.8 million EEG sessions — equivalent to 400+ years of brain data. A transformer-based architecture that decodes neural signals into actionable insights.
Jaoude et al. Deep learning models (DSN and MNet) achieved 85–86% accuracy and Kappa ~0.78 across Wake, N1–N3, and REM. Trained on 200 recordings, validated on 41. AASM-aligned.
Predicts biological brain age from at-home EEG. Identifies deviations from expected aging trajectories across diverse populations.
Migraine prediction (90% accuracy) and long COVID prediction (73% accuracy) from wearable EEG data. Published and validated.
Pre-trained models enable fine-tuning with as few as 30 participants — reducing R&D costs by up to 80% and accelerating time-to-market for new applications.
Normative database
16.8 million sessions from 122,000+ unique users across ages 18–90 with global representation. Labeled datasets include PTSD, long COVID, Parkinson's, MCI, Alzheimer's, depression, insomnia, chronic pain, migraine, and TBI.
Get started
Frequently asked questions
0 items
Your cart is empty
When you add products, they will appear here.
