What happened
Oculogica of New York makes EyeBOX, which tracks both eyes for 220 seconds while a patient watches a video move around a screen. Software applies a model to the gaze data and gives a score from 0 to 20, and a score of 10 or more is positive. In the clinical study that FDA reviewed, 282 patients with head injury were analyzed. A clinical reference standard defined concussion. Investigators revised it during the study because clinicians were often uncertain. FDA reviewers report a sensitivity of 80.4% (37/46) and a specificity of 66.1% (156/236). Neither reached the goal of a lower confidence limit above 70%. The negative predictive value was 94.5%, and the positive predictive value was 31.6%. FDA granted the De Novo request on 28 December 2018 and judged that the probable benefits outweigh the probable risks. It authorized the device only as an aid to a standard neurological assessment. The study measured accuracy and did not measure patient outcomes.
Sources
-
Regulatory document, by authors with no tie to the system.
U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo Classification Request for EyeBOX (DEN170091): decision summary. Granted 2018-12-28.
Although the primary endpoint was not met, the effectiveness analyses described above demonstrated that the probable benefit of the device outweighs the probable risk, as discussed in the Benefit-Risk Determination section below.
Details
AI system
- Workflow
-
The patient watched a video move around a screen five times while a camera tracked both eyes at 500 frames per second. A reference standard based on concussion symptom and cognitive scores defined each case.
Follow-up
- Conditions for replication
-
The patient is 5 to 67 years old and within one week of a head injury. Clinicians use the result with a standard neurological assessment and do not rely on it alone. The patient has no disorder of eye movement and is not intoxicated.
Entry details
- Sector
-
Healthcare.
- Purpose
-
Oculogica built EyeBOX to measure and analyze eye movements as an aid in diagnosing concussion within one week of a head injury.
Not yet checked: model, release, provider, role, configuration, inputs and outputs, autonomy, human oversight, AI contribution, beneficiaries, reach, distribution and access, duration, every measurement question, every money question, ownership and control, dependencies, cross-border connections, geopolitical implications, replication attempts, availability, institutional decision, grounds, action and result, success criterion, every evidence question, observed or intended, alternative assessments, related entries, other identifiers, category mappings.
Cite this record
Peter H. Vartanian. (2026). EyeBOX missed its accuracy goals but helped rule out concussion (AIBD-0784). Artificial Intelligence Benefits Database, edition of 31 October 2026. https://petervartanian.xyz/aibd/records/AIBD-0784.html (archived at https://doi.org/10.7910/DVN/AADSII, version 4.0)