Improved AI tracks moving animal neurons

Recent advances allow imaging of neurons inside freely moving animals. However, to decode circuit activity, these imaged neurons must be identified and tracked by computer. This becomes particularly difficult when the brain itself moves and deforms inside the flexible body of an organism, for example in a worm. Until now, the scientific community lacked the tools to solve the problem.

Now, a team of scientists from EPFL and Harvard has developed a pioneering AI method for tracking neurons inside moving and deforming animals. The study, now published in Natural methodswas led by Sahand Jamal Rahi from the EPFL Faculty of Basic Sciences.

The new method is based on a convolutional neural network (CNN), which is a type of AI trained to recognize and understand patterns in images. This involves a process called “convolution,” which looks at small parts of the image, like edges, colors, or shapes, at a time, then combines all of that information to make sense of it and identify objects or patterns. .

The problem is to identify and track neurons during a movie of an animal’s brain; many images must be labeled by hand because the animal appears very differently over time due to the many different body deformations. Given the diversity of animal postures, manually generating a sufficient number of annotations to train a CNN can be daunting.

To address this issue, the researchers developed an improved CNN featuring “targeted augmentation.” This innovative technique automatically synthesizes reliable annotations for reference from a limited set of manual annotations. The result is that CNN effectively learns the brain’s internal deformations and then uses them to create annotations for new postures, significantly reducing the need for manual annotations and double-checking.

The new method is versatile and can identify neurons, whether they are represented in images as individual points or 3D volumes. The researchers tested it on the roundworm Caenorhabditis elegans, whose 302 neurons have made it a popular model organism in neuroscience.

Using the improved CNN, the scientists measured the activity of some of the worm’s interneurons (neurons that connect signals between neurons). They found that they exhibited complex behaviors, such as changing their response patterns when exposed to different stimuli, such as periodic bursts of odors.

The team has made its CNN accessible, providing a user-friendly graphical user interface that incorporates targeted augmentation, streamlining the process into a complete pipeline, from manual annotation to final proofreading.

“By significantly reducing the manual effort required for segmenting and tracking neurons, the new method increases analysis throughput three-fold compared to full manual annotation,” explains Sahand Jamal Rahi. “This advance could potentially accelerate brain imaging research and deepen our understanding of neural circuits and behaviors.”