UCL scientists reconstruct videos from mouse brain activity

September 16, 2026 at 05:31 PM1 min

This digest was compiled by AI from multiple sources — links to the originals are below.

Researchers at University College London reconstructed short videos from neural activity recorded in mice, recreating what the animals saw. The study, published in eLife, used single-cell recordings from the visual cortex to generate high-quality reconstructions. The work offers a new method to compare brain representations with reality.

Key Facts

The study was led by researchers at University College London and published in eLife.
Researchers recorded activity from approximately 8,000 neurons per mouse using two-photon imaging.
The reconstruction method used a dynamic neural encoding model originally developed for the 2023 Sensorium Competition.
The algorithm started from a blank image and iteratively adjusted pixels to match measured brain activity.

Reconstruction Method

Dr. Joel Bauer of the Sainsbury Wellcome Centre at UCL led the study. The team used a dynamic neural encoding model originally developed for the 2023 Sensorium Competition. The model predicts how individual neurons respond while mice watch movies, accounting for movement and pupil diameter changes. Researchers compared predicted responses to a blank screen with actual activity during movie viewing. Neural activity was measured using a microscopic imaging method that detects localized calcium increases.

Experimental Data

Mice watched videos of natural images while neural activity in the visual cortex was recorded. Two-photon imaging captured calcium changes in approximately 8,000 neurons per mouse. The algorithm began with a blank image and iteratively modified pixels to better match the recorded brain activity. The resulting videos resembled the scenes shown to the mice, reconstructed solely from neural signals.