JMIR Mental Health Publishes Cross-Dataset Validation of Videra's Video-Based Tardive Dyskinesia Model

Videra Health’s AI can watch a short video of a patient and detect signs of tardive dyskinesia (TD), a condition involving involuntary movements of the face and body that can develop as a side effect of long-term use of certain psychiatric medications. A new study in JMIR Mental Health, co-authored by the Videra team and movement disorder specialists, put that technology through one of the toughest tests in AI research: evaluating it on an entirely new group of patients, scored by different doctors, using a different clinical rating scale than the one it learned from.

The AI’s assessments closely tracked the specialists’ judgments, correctly distinguishing patients with and without TD roughly 9 times out of 10. That’s significant because an AI tool that only works on the data it was trained on isn’t ready for the clinic. Passing an independent test like this is strong evidence the technology can be trusted with real patients, bringing routine, camera-based TD screening closer to everyday care.

Read the full paper in JMIR Mental Health.