科学家利用小鼠脑神经活动成功重构视频画面

伦敦大学学院的研究团队利用小鼠视觉皮层中单个神经元的记录，成功重构了长达10秒的视频片段1。该研究由圣斯伯里威康中心的Joel Bauer博士领导，发表在《eLife》期刊上1。

研究人员采用了钙成像方法检测单个脑细胞的活动1，并使用动态神经编码模型将神经元活动转化为视觉信息1。这一模型最初由另一研究团队在2023年Sensorium竞赛中开发1。与传统的脑成像技术不同，本研究基于单个脑细胞的记录而非fMRI信号1。研究人员通过像素相关性方法评估重构视频与原始视频的匹配度，结果显示时间差异很小1。为了验证方法的鲁棒性，团队使用了未曾出现在训练数据中的新视频进行了更具挑战性的测试1。

Researchers at University College London have successfully reconstructed a 10-second video clip based solely on neural signals recorded from individual brain cells in the visual cortex of mice 1. The study, published in eLife, demonstrates a novel approach to converting neuronal activity into visual information through a dynamic neural encoding model 1.

Led by Dr. Joel Bauer at the Sainsbury Wellcome Centre at UCL, the team employed calcium imaging to detect activity in single brain cells rather than relying on traditional fMRI brain imaging signals 1. The dynamic neural encoding model used in the reconstruction was originally developed by another research team during the 2023 Sensorium competition 1. To rigorously validate their findings, researchers tested the method on new videos that had not appeared in the training data 1. Using pixel correlation methods to assess how closely the reconstructed video matched the original, they found only minimal temporal differences 1.

This research provides new insights into how the brain processes visual information 1.