AI Revolution in Neuroscience: Precise Tracking of Neurons in Moving Animals

The AI Revolution in Neuroscience: Accurate Tracking of Neurons in Moving Animals

Researchers from EPFL and Harvard have introduced a groundbreaking AI technique designed to streamline the tracking of neurons in animals on the move. This innovation harnesses the power of a convolutional neural network (CNN) with ‘targeted augmentation,’ resulting in a substantial reduction in the need for manual annotation. As a consequence, it promises to expedite brain imaging research and deepen our comprehension of neural behaviors.

In a collaborative effort between EPFL and Harvard scientists, an AI-based method has been developed to enhance the efficiency of tracking neurons within mobile animals, significantly reducing the requirement for manual annotation. Recent advancements have made it possible to image neurons inside freely moving animals. However, to unravel circuit activity, these imaged neurons must be computationally identified and tracked. This becomes particularly challenging when the brain itself undergoes movement and deformation within the flexible body of an organism, such as a worm. Until now, the scientific community has lacked the tools necessary to address this complex issue.

Development of the AI Method for Neuron Tracking

The newly devised method relies on a convolutional neural network (CNN), a type of artificial intelligence specifically trained to recognize and interpret patterns in images. This process involves “convolution,” where the network analyzes small segments of an image, such as edges, colors, or shapes, and combines this information to identify objects or patterns.

The primary challenge lies in identifying and tracking neurons during a movie of an animal’s brain, as the animal’s appearance varies considerably over time due to numerous body deformations. Given the diverse postures of the animal, manually annotating a sufficient number of images to train a CNN can be an arduous task.

Targeted Augmentation

To overcome this challenge, the researchers developed an enhanced CNN featuring ‘targeted augmentation.’ This innovative technique automatically generates reliable annotations for reference from a limited set of manual annotations. As a result, the CNN effectively learns the internal deformations of the brain and utilizes them to create annotations for new postures, significantly reducing the need for manual annotation and verification.

This novel method demonstrates versatility by identifying neurons in various representations, whether as individual points or as 3D volumes. The researchers put it to the test using the roundworm Caenorhabditis elegans, a popular model organism in neuroscience with 302 neurons.

Using the enhanced CNN, the scientists measured the activity in some of the worm’s interneurons, which act as intermediaries between neurons. They discovered complex behaviors in these interneurons, including changes in response patterns when exposed to different stimuli, such as periodic bursts of odors.

Impact on Research

The research team has made their CNN accessible to others, providing a user-friendly graphical interface that incorporates targeted augmentation. This comprehensive pipeline, from manual annotation to final proofreading, promises to significantly reduce the manual effort required for neuron segmentation and tracking. It is estimated that this breakthrough will increase analysis throughput threefold compared to full manual annotation.

Sahand Jamal Rahi, the lead researcher, remarks, “The breakthrough has the potential to accelerate research in brain imaging and deepen our understanding of neural circuits and behaviors.”

Reference: “Automated neuron tracking inside moving and deforming C. elegans using deep learning and targeted augmentation” by Core Francisco Park, Mahsa Barzegar-Keshteli, Kseniia Korchagina, Ariane Delrocq, Vladislav Susoy, Corinne L. Jones, Aravinthan D. T. Samuel and Sahand Jamal Rahi, 5 December 2023, Nature Methods.
DOI: 10.1038/s41592-023-02096-3

Funding: École Polytechnique Fédérale de Lausanne (EPFL), Helmut Horten Stiftung, Swiss Data Science Center

Frequently Asked Questions (FAQs) about Neuron Tracking

More about Neuron Tracking

• EPFL News: “AI Revolutionizes Neuroscience with Automated Neuron Tracking”
• Harvard Gazette: “AI Method Accelerates Brain Imaging Research”
• Nature Methods: “Automated neuron tracking inside moving and deforming C. elegans using deep learning and targeted augmentation”