Clinical and scientific reference

CogniScore and BrainAge

Validation, reliability, methodology and clinical interpretation of two EEG-derived measures. Both are obtained from an eight-minute resting-state EEG recording.

For physicians, neuroscientists and other healthcare and research professionals.

Scope. BrainTrip provides a screening layer for early detection, longitudinal monitoring and informed decision-making. These measures support clinical judgement; they do not diagnose, do not treat and do not replace the clinician.

01 / Performance

Validation and reliability

Validation assesses agreement with the reference outcome. Repeatability characterises variation across repeated recordings.

Summary of validation and repeatability
Measure	CogniScore	BrainAge
Output	Cognitive function, 0–100 points	Estimated functional brain age, in years
Validation	96% specificity; 88% sensitivity At a cutoff of 50 points, against the cognitive reference classification	3.53 years mean absolute error Relative to chronological age
Repeatability	ICC 0.933 Median absolute repeat difference: 4 points	1.67 years within-person RMS variability Mean within-person SD: 1.13 years
Evaluation sample	Repeatability: 502 people, typically 1–2 weeks between recordings	Age estimation: 1,392 recordings Repeatability: 329 people
Study details	CogniScore validation and repeatability	BrainAge accuracy and variability

Sources: BrainTrip internal validation and repeatability reports. Validation and repeatability refer to distinct evaluations. Study context ↗

02 / CogniScore

CogniScore: interpretation

Measurement and clinical interpretation

CogniScore combines EEG biomarkers into a single score. Higher values indicate a pattern associated with better cognitive performance.

CogniScore summarises 393 spectral EEG biomarkers on a scale from 0 to 100. It provides an objective physiological perspective alongside the clinical history, cognitive assessment and other investigations.

A value below 50 points supports further assessment for cognitive impairment. The score does not establish the cause of impairment or distinguish dementia subtypes.

EEG and cognitive dysfunction: [1] [2] [3] [4]

Report display — illustrative example. The descriptive bands organise the report; the validation cutoff of 50 points is a separate classification threshold and lies within the reduced range. Interpret the score together with its uncertainty and clinical context.

Report interpretation

Consider the score alongside its 95% confidence interval, percentile and positive predictive value. Their definitions and interpretation are provided in the report.

THE EVIDENCEValidation, reference tests and repeatability

Validation against cognitive assessment

The reference framework combines five cognitive screening tests: MoCA, ADAS-Cog, ACE-III, Eurotest and Phototest. Measured in 502 people, the reported specificity at 50 points is 96% and sensitivity is 88% against the cognitive reference classification.

This comparison concerns cognitive impairment classification. It should not be interpreted as disease-specific accuracy for Alzheimer’s pathology.

The published comparison of the five tests provides background to this reference framework. [5]

Specificity

96%

Sensitivity

88%

One operating point at a cutoff of 50; this is not a full ROC curve.

Consistency across recordings

Repeatability was assessed using 1,372 recordings from 502 people, with two or three recordings per person, typically 1–2 weeks apart.

Single-measure ICC: 0.933
Median absolute difference, recordings 1–2: 4 points
Mean within-person standard deviation: 5.4 points
95th percentile of absolute differences, recordings 1–2: 33 points

Four points describes the typical absolute difference, not a universal limit. Some people have substantially larger changes; variability also depends on score level.

ICC: one-way random-effects, single-measure estimate accommodating unequal numbers of recordings. Methods background: [17] [18] [19] [20].

03 / BrainAge

BrainAge: interpretation

Estimated functional brain age

BrainAge expresses the age-related characteristics of an EEG in years. The difference from chronological age is the brain-age gap.

BrainAge combines 1,003 EEG biomarkers using a Bayesian inference model. Age-bias correction reduces systematic age-related estimation bias.

A positive gap means an older-appearing EEG pattern; a negative gap means a younger-appearing pattern. Neither establishes a diagnosis or an individual rate of biological ageing.

A markedly younger estimate is not necessarily favourable: altered brain activity, including patterns associated with cognitive impairment, may also lower BrainAge.

Brain-age estimation and correction: [6] [7] [9] [10]

Report display — illustrative example. Chronological age 60 years; BrainAge 62 years; gap +2 years. The scale is centred on chronological age and spans 20 years either side.

Within ±4 years

Close to chronological age

More than 4, up to 12 years

Younger or older

More than 12 years

Much younger or much older

The report’s gap p-value provides a separate measure of statistical unusualness; it is not determined by these descriptive bands alone.

THE EVIDENCEAge-estimation accuracy and within-person variability

Age-estimation accuracy

The age-estimation evaluation used 1,392 recordings across 10 test folds. Mean absolute error relative to chronological age was 3.53 years; root mean squared error was 5.08 years.

These values describe population-level agreement with chronological age. They are not an individual confidence interval or the expected difference between two recordings.

Repeat-recording variability

Among 329 people with repeat recordings, the mean within-person SD was 1.13 years. The root mean square of within-person SDs was 1.67 years, giving greater weight to people with larger variability.

The 1.67-year measure is used in the overview so that larger individual fluctuations are better represented. Neither value is a 95% limit on repeat changes. [8] [20]

Between recording estimates and age3.53 years MAE

How close is the estimated age?

Within the same person1.67 years RMS variability

How much do repeat estimates vary?

Source: internal BrainAge development evaluation with linear age-bias correction. Accuracy and repeatability describe distinct quantities and should not be treated as interchangeable.

04 / The technology

NeuroAI: acquisition and analysis

NeuroAI connects acquisition, automatic preprocessing and inference. Each stage has a distinct role in producing a usable result.

Smart Recorder
Twenty-four-channel EEG is acquired over eight minutes, comprising four minutes with eyes open and four minutes with eyes closed. Quality monitoring covers sensor contact, muscle activity, drift, reference and electromagnetic interference.
triPPPy
A 24-step automatic preprocessing workflow detects and addresses recording artefacts and standardises the EEG for analysis. Artefact correction and channel interpolation are applied where appropriate. Evaluation against expert annotations is documented in technical reports and conference contributions.
Inference Engine
Selected EEG biomarkers are calculated and combined by the product-specific models. CogniScore uses 393 biomarkers and a neural-network model; BrainAge uses 1,003 biomarkers and Bayesian inference. Demographic adjustment and age-bias correction are applied as described in the respective methodology.

BEHIND THE PIPELINEBiomarker discovery and model inference

Biomarker selection during development

Development evaluates candidate EEG biomarkers and selects combinations associated with the target outcome. A single EEG feature may carry limited information; combining complementary features can improve prediction.

Application of the selected model

During a test, the selected biomarkers are calculated and passed through a fixed model: a neural-network model for CogniScore and Bayesian inference for BrainAge. The test does not repeat biomarker discovery for each person.

General processing and spectral-method references: [11] [12] [13] [14].

05 / Clinical considerations

Factors affecting interpretation

Demographic adjustment is incorporated into the methodology. Medication, neurological conditions and current physiological state remain relevant to interpretation.

CogniScore biomarkers are adjusted for demographic factors.

AgeSexEducationHandednessHead sizeTime of recording

These adjustments reduce the influence of the listed background factors. The effects of medication and clinical conditions are not removed. BrainAge uses age as its prediction target and applies its own age-bias correction.

Medication and clinical conditions

The directions below are tendencies for interpretation, not correction factors. Effects vary with the drug, dose, condition and individual.

CogniScore

FactorLikely directionClinical interpretation

Anti-dementia treatment↑ Can raise the score

May give a higher score than cognitive performance alone would warrant, increasing the risk of a false-negative classification. Effects differ between treatments and patients. [15]

CNS depressants and sedating medicines↓ Can lower the score

Sedatives, some anxiolytics, muscle relaxants and some antiseizure medicines may lower the score, increasing the risk of a false-positive classification.

Parkinson’s disease↓ Can lower the score

EEG changes may lower CogniScore beyond what measured cognitive performance would suggest. True cognitive impairment may also coexist. [16]

Stroke or hypoxic brain injury↓ May lower the score

A lower value may reflect acquired brain dysfunction; it should not automatically be labelled a false positive or attributed to neurodegeneration.

Antidepressants and other CNS medication↕ Variable

Consider the specific medicine, dose, timing and recent changes separately from the condition being treated.

Epilepsy, schizophrenia and other neurological or psychiatric conditions↕ Variable

Clinical conditions and their treatments may alter EEG patterns. Direction and magnitude cannot be inferred from the diagnosis alone.

Alcohol and other psychoactive substances↕ Variable

Sedating effects can lower values; other effects vary. The history covers alcohol in the last 24 hours and psychoactive substances in the last 48 hours. These are history windows, not universal wash-out periods.

Mood, motivation and cooperation

BrainTrip’s clinical interpretation guidance describes CogniScore as relatively robust to depression, anxiety, mood, motivation and task cooperation. Resting EEG does not depend on performing a cognitive task, although cooperation is still needed for a usable recording. Medicines used to treat these conditions may affect the score.

BrainAge: state-dependent influences

BrainTrip’s interpretation guidance describes BrainAge as more dependent on psychological and physiological state than CogniScore. Mood, anxiety, stress, sleep, fatigue and relaxation can influence the EEG pattern.

Medication and disease

CNS-active medication, neurological and psychiatric conditions, alcohol and other psychoactive substances may shift the estimate. A single reliable direction cannot be assigned to every drug or diagnosis.

Interpreting a change

An older- or younger-appearing EEG may reflect a change in functional state. It should not automatically be interpreted as accelerated ageing, rejuvenation, or treatment benefit.

The relative psychological-state sensitivity is clinical interpretation guidance; a quantitative head-to-head comparison is not provided here.

The boundary, stated plainly

These measures are a screening layer: early detection, longitudinal monitoring and better-informed decisions. They do not diagnose, do not treat and do not replace clinical judgement. A result outside the expected range routes into the institution’s own referral pathway.

RELEVANT HISTORYWhat to review alongside the report

Conditions and clinical context

Subjective cognitive complaints; family history of dementia; a dementia diagnosis and its type; stroke or heart attack in the past 12 months; Parkinson’s disease; epilepsy; other neurological conditions; depression; schizophrenia; and other psychiatric conditions.

A recent heart attack is contextual information, especially if accompanied by hypoxia or neurological complications; it does not imply a uniform score change.

Treatments and recent exposures

Anti-dementia treatment, antidepressants, other CNS-active medication, alcohol within 24 hours and other psychoactive substances within 48 hours. Record the actual medicine, dose, timing and recent changes where available.

Treatment-related physiological changes are not simply recording artefacts. A score change alone does not establish treatment efficacy. Medication changes or withdrawal require the prescribing clinician’s judgement.

Product-specific directional guidance: BrainTrip clinical interpretation materials. External studies support the underlying EEG effects; they do not independently validate each product-specific effect.

06 / Scientific evidence

Scientific evidence and references

Evidence is grouped into product evaluations, conference contributions and peer-reviewed methodological and clinical literature.

Product evidence

Validation and repeatability

Internal CogniScore validation and repeatability analyses, BrainAge development evaluations, and clinical interpretation guidance underpin the product figures on this page.

Evaluation context

CogniScore classification statistics refer to the cognitive reference evaluation at a cutoff of 50. Repeatability comes from a separate repeated-recording analysis.

BrainAge metrics correspond to the documented 10-fold evaluation with linear age-bias correction. Evaluation configuration and release version must remain paired when performance figures are updated.

Internal reports are distinct from peer-reviewed publications. This page does not distribute the underlying datasets or internal documents.

Research contributions

Publications and conference contributions

BrainTrip’s research record includes the cognitive screening test comparison and conference work on EEG processing, cognitive screening and brain-age estimation.

Research record

2022 · Society for NeuroscienceValidation of an EEG-based dementia screening test
2023 · Society for NeuroscienceFully automatic EEG preprocessing
2024 · MOBI, FENS and OGBMTConference contributions to EEG methodology and cognitive assessment
2024 · Peer-reviewed publicationComparison of five cognitive screening tests [5]
2025 · Society for Neuroscience abstractEEG-based brain-age estimation

THE WIDER LITERATUREReferences (20 publications)

These references support the scientific background and methods. They should not be read as 20 independent validations of CogniScore or BrainAge. Studies using sleep EEG, imaging or other processing pipelines have different protocols and models.

Find a reference20 references

DOI registration and bibliographic metadata checked in October 2026. Full-text access depends on the publisher.