Neurons in the brain, the basis of thought, connect to each other using small surface branch-like structures known as dendritic spines. Now, scientists at Columbia University’s Zuckerman Institute have devised powerful new artificial intelligence-powered software that can automatically map these dendritic spines on photographs of neurons, and the researchers are making the tool available for free.
“Dendrite spines are usually the first sites involved in neurodegenerative diseases such as Alzheimer’s disease and Parkinson’s disease,” Sergio Bernal García, a graduate student in Dr. Frank Pollu’s lab and lead author of a new study detailing this work, said in the journal Oct. 20. cell report method. “So it’s really important to understand more about them.”
Dendritic spines are currently counted mostly manually. Painstakingly analyzing images of hundreds of neurons can take weeks or months. The new tool, called RESPAN (Repair Enhanced Spine and Neuron Analysis), “takes just a few minutes on the computer,” Bernal García said.
RESPAN automatically identifies dendritic spines and measures their volume, length, and surface area. This software can display the position of spines on cells and calculate their distance from the central part of the cell, and can do this in live animals. It also provides multiple optional image restoration steps to help analyze particularly difficult images, as well as a way for users to train the software on their own datasets.
RESPAN not only outperformed manual analysis, it also proved to be more accurate and detect fewer false positives and negatives than previous neuron analysis tools. “By using our freely available tools, researchers can significantly improve the consistency and reliability of their results, helping to address the reproducibility crisis in biomedical sciences,” said Luke Hammond, senior and corresponding author of the study, former director of the Cell Imaging Platform at the Zuckerman Institute and current director of the Division of Quantitative Imaging in the Department of Neurology at The Ohio State University Wexner Medical Center.
The researchers aimed to make RESPAN as user-friendly as possible. “Scientists often revert to manual approaches because the software packages that exist for this task lack functionality or have limited precision when analyzing difficult images,” Hammond said. “Importantly, users do not need to know any coding to use RESPAN and there are YouTube tutorials to guide users through each step.”
Researchers hope to make new discoveries with a new tool that allows them to quickly and automatically map every dendritic spine on a neuron. “By spatially mapping all the spines on a neuron, we can now reveal whether certain locations are more susceptible to disease and investigate whether spines in different regions have distinct molecular signatures,” Bernal Garcia said.
RESPAN can run on a PC or laptop with an NVIDIA GPU. This software is open source. That means other people are free to tinker with it as they like. “We encourage the community to adapt and improve RESPAN,” Bernal-Garcia said.
reference: Bernal-Garcia S, Schlotter AP, Pereira DB, Recupero AJ, Poliu F, Hammond LA. A deep learning pipeline for accurate and automated reconstruction, segmentation, and quantification of dendritic spines. cell report method. 2025;5(10):101179. doi: 10.1016/j.crmeth.2025.101179
This article has been reprinted from the following material: Note: Materials may be edited for length and content. Please contact the citation source for details. You can access our press release publishing policy here.
