
Starting from just a tiny piece of brain tissue, scientists at Baylor College of Medicine can now reconstruct the complete neural framework. They've simplified many laborious parts of this procedure through the creation of a specialized tool named NEURD, which stands for "NEURal Decomposition." This innovative software accelerates the identification and rectification of errors within the datasets detailing neuronal connectivity—essentially mapping out how different regions communicate—and thus facilitates groundbreaking findings.
In an article published in the latest version of Nature In their work, researchers detail how NEURD streamlines the preparation of complex datasets for multiple subsequent analyses—by checking accuracy, refining them, and offering simpler methods to query and label the information with understandable attributes. Additionally, they demonstrate through the use of NEURD, this refined and labeled dataset becomes more accessible for exploration, leading to several novel discoveries regarding the brain.
The main objective of this partnership was to elucidate the significance of neural wiring within the brain, exposing the biological ‘key ingredient’ responsible for enabling the remarkable problem-solving abilities of our brains, capabilities that still elude today’s most advanced machine learning technologies,” stated Dr. Jacob Reimer, an assistant professor of neuroscience at Baylor University, a founding member of the Center for Neuroscience and AI, and lead researcher on the study.
MICrONS
The collaboration mentioned by Reimer is The MICrONS Project, which comes with a massive electron microscopy dataset providing an unparalleled view into the structure and functionality of the mouse brain at the millimeter scale. This initiative lasted for seven years and brought together over 150 researchers from top global institutions.
The international team developed the most extensive and comprehensive neural circuit blueprint of a mammalian brain up until now using just one cubic millimeter of a mouse’s visual cortex. This intricate model encompasses over 523 million synaptic connections, numerous axonal pathways extending for several miles, along with more than 200,000 individual cells. What sets this particular wiring diagram apart is its inclusion of both the practical attributes of neurons—their electrical activity—as well as their physical structure and interconnectivity details. By doing so, it offers an unparalleled insight into neuron interaction and information processing within networks, thereby enhancing our understanding of these covert dialogues occurring inside the brain.
The MICrONS dataset encompasses not just high-resolution anatomical images, but also incorporates real-time functional data gathered from the identical set of cells. Although a limited number of prior datasets have combined functionality with structural information, they did so on a significantly smaller scale.
An innovative imaging technique was employed to document this activity. As a graduate student in the Tolias Lab at Baylor—now affiliated with Stanford University—Reimer collaborated with a group of scientists who recorded the reactions of these neurons over several weeks while the mice encountered various visual stimuli, including films like "Mad Max."
Scientists from the Allen Institute for Brain Science handled the acquisition of anatomical electron microscopy data from dissected tissues, while partners at Princeton University tackled the task of reconstructing the extensive (petabyte-level) electron microscopy anatomical volume. This process involved aligning all the images and dividing them into segments to generate a three-dimensional representation of every neuron along with their connecting pathways.
This method enables researchers to examine the brain’s architecture and functionality simultaneously, offering an unprecedented perspective on how neurons within the brain handle sensory data. Having established his laboratory at Baylor, Reimer has joined forces with the Tolias lab as well as other partnering research centers to investigate the relationship between functional aspects and structural elements of the brain.
Enter NEURD
To enhance the precision of the data and extract valuable insights, the researchers from Baylor together with their partners created NEURD—a sophisticated computational instrument designed to review and cleanse the initial data sets. This ensures the information’s reliability for scientific applications.
NEURD streamlines the proofreading procedure by detecting and rectifying mistakes within neural networks, refining and marking the 3D architecture. To assess the reliability of this automatic proofreading method, researchers conducted thorough validations to gauge the precision and recall of the corrections made. This ensures that following examinations can be executed with maximum accuracy.
“NEURD addresses the issue of automated neuron proofreading by dividing it into a series of less complex tasks that align closely with how humans generally observe neurons. By starting with recognizing elements such as spines, we can better determine cell types, which then aids in differentiating between axons and dendrites—thus making the entire proofreading process more structured and easier,” explained Dr. Brendan Celii, who led the research and was previously affiliated with Baylor and Rice Universities before joining the Johns Hopkins University Applied Physics Laboratory.
Expanding upon current software tools used for generating 3D mesh representations of neurons, NEURD operates via automated identification of crucial elements such as neuron cell nuclei (somas), minuscule extensions known as spines found on neurons, along with precise delineation of neuronal axons and dendrite branches. This suite of capabilities is intended to function effectively across various datasets, enabling scientists to collect uniform information applicable to multiple experimental conditions.
Reimer stated that proofreading is just one part of NEURD, with the additional facet being to carry out scientific research.
As increasingly precise diagrams of neural pathways are generated across different species and brain areas, NEURD will play an essential role in enhancing our comprehension of neural networks. Although this field remains in its infancy, the potential impact on human well-being is significant. Researchers aim to pinpoint irregularities within particular neuronal circuits that might result in ailments such as Alzheimer’s, Parkinson’s, or autism spectrum conditions. With deeper insights into these issues at a cellular scale, therapies may be devised that address the underlying sources of neurodegenerative diseases.
Reimer and his team outline various methods for utilizing the annotated 3D graphs generated by NEURD. These techniques include classifying distinct types of neurons followed by determining the likelihoods of synaptic connections among these classifications.
He suggested we can inquire about the connections among various classes or differing levels of spinal density. These details have been calculated beforehand, hence you only need to come up with queries; the information awaits discovery within the dataset," he mentioned. "This study encompasses an astounding magnitude, despite originating from a mouse’s neural network. The ability to analyze such extensive data will provide us with unprecedented insights into how neurons connect in our own brains.
Although additional efforts are necessary and scientists remain distant from completely charting the intricacies of the human brain, the advancements achieved so far have been revolutionary.
"We are stepping into a new epoch in neuroscience, where machine learning can merge with biological comprehension to unravel the enigmas of the brain. We aim for NEURD to assist in getting additional researchers involved in deriving knowledge from these extensive datasets," Reimer stated.
More information: Celii, B., et al, state that NEURD provides automatic proofreading and feature extraction capabilities for connectomics. Nature (2025). DOI: 10.1038/s41586-025-08660-5 . www.nature.com/articles/s41586-025-08660-5
Furnished by Baylor College of Medicine
The tale was initially released on Medical Xpress . Subscribe to our newsletter For the most recent science and technology news updates.