In a groundbreaking development, researchers have unveiled three artificial intelligence tools capable of tracking individual neurons in real-time, even in the most elusive of subjects like the “wiggly” roundworm or the deforming jellyfish. This advancement, announced by the Picower Institute at MIT, marks a significant leap in neuroscience, offering automated solutions to the longstanding “alignment and annotation” bottleneck.
The newly developed tools—BrainAlignNet, AutoCellLabeler, and CellDiscoveryNet—can identify and track cells with up to 99.6% accuracy, dramatically reducing the manual labor previously required for such tasks. This innovation not only accelerates research but also opens new avenues for understanding the nervous systems of living, behaving organisms.
The Challenge of Tracking Neurons
Understanding the connection between brain activity and behavior is a major goal in neuroscience. However, tracking neurons in live animals is notoriously difficult due to the constant movement and deformation of the organism, which shifts the position of brain cells. The challenge is particularly pronounced in simple, transparent lab animals like the roundworm C. elegans and the jellyfish C. hemisphaerica.
According to Steven Flavell, senior author of the study and associate professor at The Picower Institute, “In a live behaving animal, we can now keep track of neurons over time and even determine the exact identities of most neurons. This is essential for our goal of relating brain activity to behavior.”
Introducing the AI Tools
BrainAlignNet
BrainAlignNet excels in tracking cells through long video series with remarkable speed and precision. It operates 600 times faster than previous methods, achieving single-pixel accuracy of 99.6% compared to ground truth. This tool rigorously addresses the alignment problem, ensuring that researchers can follow the same neuron across different frames.
AutoCellLabeler
AutoCellLabeler is designed to identify specific cell types within images. It uses a color-coding system and can achieve 98% accuracy, even with minimal training data. This tool is particularly effective in identifying neuron types such as the “NSM” neuron, significantly reducing the time and expertise required for manual annotation.
CellDiscoveryNet
CellDiscoveryNet offers a unique capability to identify and cluster cell types across different animals without human supervision. Its performance matches that of expert human labelers, making it a valuable asset for labs dealing with diverse and complex datasets.
Broader Implications and Applications
While these tools were initially developed for the roundworm C. elegans, they have already been applied successfully to the more complex nervous system of the C. hemisphaerica jellyfish. This adaptability suggests a wide range of potential applications in other biological contexts.
Brady Weissbourd, a co-author of the study and assistant professor at MIT, highlighted the significance of these tools in his research. “They call it a jellyfish for a reason. Any part of it can move relative to any other part of it. The tool helped us to register our videos to be able to extract neural activity from them,” Weissbourd explained.
Future Directions and Challenges
Despite these advancements, researchers acknowledge that there is still more work to be done. Weissbourd is currently working on labeling all the cells in the jellyfish, and developing a microscope capable of imaging the jellies as they swim freely. The potential for these tools to be generalized to other contexts requiring alignment and annotation of dense heterogeneous cell types in complex tissues is immense.
Funding for this research came from various prestigious sources, including the National Institutes of Health, the National Science Foundation, and The Howard Hughes Medical Institute, underscoring the importance and potential impact of this work.
The study, titled “Deep neural networks to register and annotate cells in moving and deforming nervous systems,” was published in eLife and represents a collaborative effort among researchers from MIT and other institutions.
As neuroscience continues to evolve, the integration of AI tools like BrainAlignNet, AutoCellLabeler, and CellDiscoveryNet will undoubtedly play a crucial role in advancing our understanding of complex nervous systems and their behaviors.
