01 / Overview
Multiple signals.
A careful interpretation.
Eye movements, vestibular responses, reaction time and cognition offer complementary ways to study neurological function after mild traumatic brain injury (mTBI). Published studies support continued investigation—but results vary by task, device, population and time since injury. [3][4][7][9][11]
What the evidence supports
Several studies report measurable group differences and useful discrimination with particular OVRT and eye-tracking protocols. Multimodal clinical assessment provides the context in which such findings should be interpreted. [1][2][3][6]
What remains unresolved
Reproducibility across settings, added value over existing assessment, and reliable individual-level interpretation remain central questions. Strong model-development results do not guarantee performance in new participants. [7][9]
Evidence boundary. This review summarizes external literature. It does not report clinical validation of Cerebro or Neocognitron. Neocognitron is investigational and is not cleared for diagnosis or independent treatment, disposition, or return-to-activity decisions.
02 / Clinical context
Measurements inform
the clinical picture.
The 2019 American Medical Society for Sports Medicine position statement provides a clinical foundation for evaluating sport-related injury. The later Amsterdam consensus, published in 2023, emphasizes multimodal and serial evaluation. History, symptoms, examination, cognition, balance and gait must be interpreted together; a single measurement should not substitute for that assessment. [1][2]
In this review, mTBI means mild traumatic brain injury. “Concussion” remains in original study titles and where it identifies the population studied. Sports-related findings are not automatically generalizable to every mechanism, age group or severity of traumatic brain injury.
- OVRT
- Oculomotor, vestibular and reaction-time measures: a family of tasks, not one standardized diagnostic test.
- VOMS
- Vestibular/Ocular Motor Screening: a clinical screen emphasizing symptom provocation and near-point convergence, distinct from instrumented eye-movement measurement.
- AUC
- Area under the receiver operating characteristic curve: a measure of discrimination across thresholds, not a percentage accuracy or evidence of clinical utility.
- Reliability ≠ validity
- A repeatable measurement can still fail to identify the intended condition. Agreement, diagnostic discrimination and improvement in patient care are separate questions.
03 / Selected evidence
Oculomotor, vestibular & reaction-time measures
OVRT batteries combine eye-movement tasks, vestibular challenges and response timing. Their results depend on the specific hardware, stimulus protocol and analysis—not simply on measuring several domains.
Early evidence from emergency-department cohorts
Two cohorts · 50 mTBI participants and 100 controls per cohort · assessed within 6 days
Reported finding
A selected three-test combination reported 89% sensitivity and 95% specificity for separating clinically diagnosed mTBI from controls.
Interpretation & limits
These are study-specific results from an infrared video-oculography and rotational-chair protocol. They do not establish the performance of a different device, task battery or population.
Patterns change during early recovery
106 participants with mTBI · 300 controls · presentation and 1- and 2-week follow-up
Reported finding
Six measures from five tests classified the initial session with 84.9% sensitivity and 97.0% specificity. The pattern of abnormalities evolved across visits.
Interpretation & limits
A longitudinal change in a test is different from a validated recovery endpoint. This work extends earlier OVRT research; it should not be counted as wholly independent replication of that program.
Eye movements and response timing in sport
Sports-related injury cohort · video-oculography
Reported finding
The study reported differences in pursuit initiation, reaction time, saccade velocity and optokinetic responses. Some optokinetic findings were associated with longer symptom duration.
Interpretation & limits
The findings support studying several functional responses together. Sports-cohort classification models require separate validation before use in emergency or general clinical populations.
04 / Selected evidence
Eye tracking & computational classification
Promising discrimination and limited or negative results coexist. The distinction between model development and evaluation in held-out participants is essential.
An eye-movement classifier with a separate validation dataset
Prospectively assessed injured participants and controls
Reported finding
The selected classifier reported an AUC of 0.878, a cross-validated AUC of 0.852 in CT-negative participants, and an AUC of 0.831 in a validation dataset.
Interpretation & limits
The result applies to that classifier, reference definition and sample. AUC describes discrimination across thresholds; it is not the percentage of patients diagnosed correctly.
A direct comparison exposes the validation gap
Three commercial devices · 49 acute mTBI participants analyzed · 34 chronic mTBI participants · 119 controls recruited
Reported finding
Acute-group AUCs ranged from 0.690–0.950 in the 80% development sample and 0.600–0.750 in the 20% validation sample. Chronic-group validation AUCs were 0.490–0.571.
Interpretation & limits
A strong development result may not persist in held-out data. Quality exclusions, small validation samples and device-specific testing problems constrain interpretation.
A multi-task eye-tracking approach
20 recent mTBI participants · 40 with persisting symptoms · 55 controls
Reported finding
The multi-task battery showed moderate-to-excellent repeatability across reported measures. Three-group machine-learning classification yielded an AUC-ROC of 0.82; reported balanced accuracy for the mTBI class was 0.66.
Interpretation & limits
This is exploratory classification evidence. Small clinical groups and validation design limit transfer to new clinical settings; repeatability of a metric does not by itself establish diagnostic validity.
Limited discrimination in an emergency setting
Prospective exploratory ED study · eye tracking with virtual-reality glasses · early assessment
Reported finding
Across 52 analyzed parameters, only two blink-related measures showed limited discriminatory value. The authors did not recommend the tested implementation for ED diagnosis in its current form.
Interpretation & limits
Feasibility is not diagnostic utility. Task design, manual processing, reference classification and the early injury window matter; favorable results from another platform cannot resolve these limitations.
Symptoms and measured function can follow different trajectories
Collegiate athletes · acute and symptom-free assessments
Reported finding
Some oculomotor abnormalities persisted when athletes no longer reported symptoms.
Interpretation & limits
The small sample, lack of an a priori power analysis and exclusions for incomplete or poor-quality data limit generalization. Persistent differences do not establish an individual return-to-sport threshold.
05 / Selected evidence
Cognition, vestibular screening & clinical context
Rapid naming, symptom provocation and instrumented eye tracking measure different constructs. Their findings can inform a multimodal assessment, but their scores and diagnostic claims are not interchangeable.
Rapid number naming: sensitivity with a specificity trade-off
97 injured NCAA Division I athletes · 97 matched controls · within 48 hours
Reported finding
For any increase in King–Devick completion time from baseline, sensitivity was 81% and specificity 56%. AUC was 0.75 (95% CI 0.68–0.82).
Interpretation & limits
A change from an individual baseline can support further assessment, but false positives and false negatives remain. These data are not evidence for a stand-alone rule-out test or a Cerebro cognitive score.
Multi-domain symptom provocation and time to return to play
Retrospective clinic cohort · 928 athletes
Reported finding
Athletes positive on more than three VOMS domains returned to play, on average, 12.81 days later than those positive on three or fewer. Domain count explained approximately 10.8% of variation in return time.
Interpretation & limits
This is an association, not a causal effect or an individual recovery forecast. VOMS records symptom provocation; it is not equivalent to direct measurement of eye or vestibular physiology.
Vision disorders in adolescents with persisting symptoms
108 adolescents · symptoms beyond 28 days · comprehensive vision examination as comparator
Reported finding
67 participants (62%) had a diagnosed vision disorder. A multivariable model that included an eye-tracking BOX score achieved an AUC of 0.7637.
Interpretation & limits
The outcome was a vision disorder within an already symptomatic group—not acute mTBI diagnosis in the general population. The reported AUC belongs to a multivariable model, not the eye-tracking score alone.
Whiplash illustrates the challenge of specificity
15 participants with chronic whiplash-associated disorder · 15 matched controls
Reported finding
The whiplash group showed less efficient performance in 31 of 34 ocular variables and differences on several assessments used in post-injury care.
Interpretation & limits
Similar findings can occur in another clinical population. This small cross-sectional study does not establish that participants had an unrecognized mTBI, or that an abnormal eye-movement result identifies a specific cause.
Clinical examination remains central
Pediatric sideline assessment · instructional video article
Reported finding
The article describes clinical recognition and examination, including vestibular and ocular screening, in pediatric sport.
Interpretation & limits
This is an educational clinical resource, not a diagnostic-accuracy trial. It provides workflow context and should not be assigned the evidentiary weight of a prospective validation study.
06 / Selected evidence
Delivery, measurement quality & portability
A change in platform changes the measurement problem. Virtual delivery, infrared tracking and a front-facing camera each require evidence appropriate to their own implementation.
Virtual examination: reliability varies by measure
60 people with acquired brain injuries of differing types and severities · virtual versus in-person assessment
Reported finding
Interrater agreement for VOMS symptom change was high (κ = 0.93; 95% CI 0.83–1.00), while some cervical measures performed poorly. Several estimates had wide confidence intervals.
Interpretation & limits
Agreement with an in-person assessment is not accuracy for diagnosing mTBI. The mixed injury population and measure-specific uncertainty argue against treating the entire virtual battery as uniformly validated.
Pavilionis et al. · online 2025 / issue 2026
Source [15]Standardizing VOMS delivery with virtual reality
Division I athletes · post-injury groups of 29 prototype and 22 VR participants · within 72 hours
Reported finding
The comparison found no statistically significant post-injury differences in total symptom provocation or near-point convergence between the delivery methods.
Interpretation & limits
The authors identified the comparison as underpowered. Failure to detect a difference does not demonstrate equivalence, and these symptom-based findings do not validate an automated physiological measurement.
Smartphone gaze estimation as a measurement opportunity
Technical eye-tracking study · calibrated smartphone camera · controlled viewing conditions
Reported finding
Personalization using approximately 30 seconds of calibration produced a mean on-screen gaze error of about 0.46 cm under the study conditions. Accuracy varied with head pose and viewing distance.
Interpretation & limits
Spatial gaze accuracy is not validation of saccade timing, vestibulo-ocular reflex measurement or mTBI detection. Clinical translation requires device-, task- and population-specific measurement studies.
07 / Implications for Cerebro
A rationale for research.
A responsibility to validate.
Cerebro’s research question is whether relationships among functional signals provide information beyond individual measurements. The literature motivates that question; it does not establish that Cerebro’s combination, hardware or models answer it.
Dynamic Response Behavior (DRB)
Dynamic Response Behavior (DRB) describes how neurocognitive responses unfold over time—how they adapt to task demands, vary across repeated attempts, and relate across ocular, vestibular, reaction-time, cognitive, and motor domains.
This research concept guides Cerebro’s investigation of relationships across signals. It is not a validated clinical score or diagnostic finding. The cited studies inform the underlying research rationale; they do not establish or validate DRB as a unified construct, and their authors may use different terminology.
Our interpretation of these studies leads to four priorities for a defensible evaluation:
Establish measurement validity.
Evaluate calibration, timing, repeatability and failure rates for each intended device and task. Prespecify quality thresholds and report excluded or incomplete assessments.
Test the intended clinical use.
Define the population, injury window, comparator and role of the result. Use prospective evaluation with blinded clinical adjudication and an appropriately representative sample.
Separate development from validation.
Lock the model and decision threshold before evaluation in independent participants. Report uncertainty, sensitivity, specificity and subgroup performance; prevent participant-level data leakage.
Demonstrate added value.
Compare the multimodal approach with individual measures and standard assessment. Investigate whether any performance gain translates into useful, safe clinical decisions.
Regulatory evidence belongs to a specific device.
The FDA record for EyeBOX EBX-4.1 describes an eye-movement assessment aid used with a standard neurological assessment in patients aged 5–67. It is an example of a defined intended use—not evidence of clearance, equivalence or a predetermined regulatory pathway for Cerebro. [19]
08 / Scope & methods
How to read this review.
This is a curated narrative synthesis prepared by Cerebro NeuroTech, Inc. It draws on Cerebro’s December 2025 OVRT review, the supplied 2025 study collection, and the AMSSM position statement, with the Amsterdam consensus added for subsequent clinical context. References and selected reported results were checked against available publisher or PubMed records for this web edition.
The emphasis is oculomotor and vestibular measurement, reaction time and brief cognitive assessment in acute mTBI. Selected studies of persisting symptoms, whiplash, virtual examination and smartphone eye tracking are included to examine generalizability and measurement limitations. They are identified separately from acute diagnostic studies.
This is not a systematic review or meta-analysis. No exhaustive database search, duplicate screening or formal risk-of-bias scoring was performed for this edition. Study selection is limited to the supplied collection and contextual sources; it should not be interpreted as complete coverage through the update date. Summaries use available abstracts and full texts and do not reproduce the original articles.
Results have not been pooled across different tests or populations. Where confidence intervals are provided, they belong to the cited study. Publication years distinguish first online publication from later journal issue dates where relevant.
Disclosure: Cerebro is developing the investigational technology discussed in the research implications. This company-authored synthesis is not independently peer reviewed, a clinical practice guideline, or regulatory advice. Readers should consult the original articles for complete methods, funding and competing-interest disclosures.
09 / References
Follow the evidence.
Original publication titles are retained. Links lead to the publisher, PubMed, PubMed Central or FDA record.
