To understand how the brain controls movement, researchers often observe small, transparent animals like roundworms. They use a process where neurons glow when they are active. This glowing is known as fluorescence. By watching these patterns of light, scientists can see which parts of the brain are working during different behaviors. However, because these animals are soft and move constantly, their brains stretch and warp. This makes it very difficult for humans to keep track of which glowing dot is which over a long period.
Researchers at MIT have developed three new computer programs to solve this problem using artificial intelligence (AI). These programs use neural networks. The first program is called BrainAlignNet. It handles a task called alignment, which involves tracking a single cell as it moves from one spot to another in a video. This software is 600 times faster than previous methods and is almost 100 percent accurate.
Automation overcomes research bottlenecks
The second tool, AutoCellLabeler, identifies the specific type or name of each neuron. This process of naming parts of an image is called annotation. While this tool requires some initial human training, it can accurately identify cells even in images with less detail. The third tool is called CellDiscoveryNet. It can identify and group similar cells across different animals without any human training or supervision.
Before these tools were created, experts had to spend hundreds of hours naming cells by hand, which was a slow and expensive process. Now, the software can do this work automatically. These tools are also useful for studying other creatures, such as jellyfish. Because a jellyfish can move any part of its body independently, tracking its neurons was once nearly impossible. These new AI systems allow scientists to gather data from these complex movements easily. This technology provides a new way for laboratories to study the brain without having to choose between speed and accuracy.
This study is published in eLife.