Вештачка интелигенција

Groundbreaking AI Technique Enables Tracking of Neurons in Moving and Deformed Animals

Summary

Recent research by scientists from the Federal Institute of Technology in Lausanne (EPFL) and Harvard University presents a revolutionary method for tracking neurons in moving and deformed animals. Published in the journal Nature Methods, this study represents a significant advancement […]

Groundbreaking AI Technique Enables Tracking of Neurons in Moving and Deformed Animals

Recent research by scientists from the Federal Institute of Technology in Lausanne (EPFL) and Harvard University presents a revolutionary method for tracking neurons in moving and deformed animals. Published in the journal Nature Methods, this study represents a significant advancement in neurological research.

Instead of manually labeling images, scientists have developed an intelligent method based on convolutional neural networks (CNN) that automatically synthesizes reliable reports for reference points from a limited set of manually labeled images. This approach allows CNN to efficiently learn deformations within the brain and use them to generate reports on new positions, significantly reducing the need for manual labeling and double-checking.

Additionally, the new method is adaptable and can identify neurons whether they are represented as individual points or 3D volumes. It was tested on the nematode worm Caenorhabditis elegans, whose 302 neurons make it a popular model in neuroscience research.

Using the improved CNN, scientists tracked the activity of some interneurons in the worm and discovered complex behaviors, such as changes in response patterns to different stimuli, like periodic odor cues.

Furthermore, the researchers ensured that their method is accessible to other scientists by providing a user-friendly graphical interface that integrates “targeted amplification,” reducing the process from manual labeling to final verification to a simple and comprehensive step.

This groundbreaking approach reduces the time and effort required for studying neurons, enabling three times greater efficiency in analysis compared to complete manual labeling. Scientists hope that this technique will accelerate research in brain imaging and deepen our understanding of neural circuits and behavior.

FAQ