Cheese3D Discovery Platform Uses Multidimensional AI Vision to Quantify Subtlest Facial Expressions in Mouse Models

CSHL researchers develop Cheese3D, a 3D camera system and AI that quantifies facial movements to study brain function and monitor anesthesia depth.

By: AXL Media

Published: Apr 27, 2026, 6:42 AM EDT

Source: Information for this report was sourced from EurekAlert

Decoding the Language of Facial Behavior

While humans intuitively grasp the spectrum of emotion in a face, science has historically struggled to quantify the nuanced relationship between facial movement and brain activity. At Cold Harbor Spring Laboratory (CSHL), Assistant Professor Helen Hou and her team have introduced a transformative discovery platform named Cheese3D to bridge this gap. Published in Nature Neuroscience, the system utilizes advanced computer vision to track even the most minute shifts in expression. By creating a reliable, automated way to measure these behaviors, researchers can now move beyond subjective observations toward a methodical interpretation of how the brain orchestrates facial muscle tone.

Engineering a Multi-Perspective Vision System

The unique cone-shaped anatomy of a mouse face required a specialized hardware solution to capture high-fidelity data. The Hou lab, in collaboration with CSHL’s Core Facilities, engineered a rig consisting of six miniature cameras that simultaneously record a mouse’s face from multiple angles. Machine learning models then act as an automated film editor, compiling these disparate perspectives into a cohesive 3D map of facial behavior. This comprehensive view ensures that no movement is missed, providing a dataset far richer than what could be achieved with a single-camera setup.

Validating Accuracy Through Anesthesia Monitoring

To demonstrate the system’s precision, researchers applied Cheese3D to mice undergoing anesthesia, a state where muscle tone changes are critical but subtle. The AI-driven platform successfully measured the depth of the "awake" or "asleep" state by analyzing facial cues alone. In collaborative tests with CSHL’s Borniger lab, Cheese3D matched the accuracy of electroencephalography (EEG), the current gold standard for monitoring brain activity. Remarkably, the system achieved this level of detail non-invasively, allowing for constant monitoring without the need for traditional, intrusive sensors.