Related Experiment Video
Updated: May 16, 2026
05:49
An Automated Squint Method for Time-syncing Behavior and Brain Dynamics in Mouse Pain Studies
Published on: November 1, 2024
Cross-Dataset Evaluation of an Automated Video-Based Model for Detecting Tardive Dyskinesia Using the Clinician's
Richard Trosch1, Anthony Sterns2, Bradley Grimm3
1William Beaumont School of Medicine, Oakland University, Farmington, MI, United States.
JMIR Mental Health
|May 14, 2026
Summary
An AI model accurately detects tardive dyskinesia (TD) using videos, correlating well with clinician scores. This technology shows promise for improving TD diagnosis in clinical settings.
Area of Science:
- Neurology
- Medical Imaging
- Artificial Intelligence
Background:
- Tardive dyskinesia (TD) is a prevalent movement disorder linked to antipsychotic use.
- Current detection methods in routine care are inconsistent despite available rating scales.
Purpose of the Study:
- Evaluate an AI-powered, video-based model for TD detection.
- Compare automated assessments with clinician-rated scales (AIMS, CTI) for reliability and accuracy.
Main Methods:
- Analyzed 69 videos using the TDtect visual transformer algorithm.
- Correlated automated predictions with clinician ratings (AIMS, CTI) using Pearson correlation.
- Assessed predictive accuracy with area under the curve (AUC) metrics.
Main Results:
- The AI model demonstrated strong correlation with AIMS scores (r=0.717) and high diagnostic accuracy (AUC 0.854, improved to 0.900).
- Highest reliability was observed for movements of the tongue, lips, and jaw.
- Functional CTI components showed weaker correlations due to their subjective nature.
Conclusions:
- AI-driven TD detection shows potential for broad clinical application across various video protocols.
- Further validation is required, with future refinements aimed at enhancing accuracy, especially for functional impact prediction.
