Researchers at SLAC National Accelerator Laboratory have developed SPINE, a machine-learning tool that reconstructs neutrino interactions from detector data, improving the efficiency of experiments like ICARUS and DUNE. SPINE recovered 60% more usable neutrino interactions than conventional methods, aiding studies on sterile neutrinos.
SPINE’s Machine-Learning Approach
SPINE, short for Scalable Particle Imaging with Neural Embeddings, uses hierarchical machine learning to analyze data from liquid argon detectors. Unlike traditional algorithms that chain narrow tasks, SPINE combines multiple models to spot features, identify particles, and reconstruct interactions step by step. This approach improves transparency and precision, as noted by SLAC scientist François Drielsma: We didn’t try to go, in one step, from the raw detector data all the way to these high-level physics observables.
The tool’s design allows it to handle complex signals from neutrino collisions, which produce overlapping particle tracks and background noise. By extracting progressively higher-level information from detector data, SPINE reduces systematic uncertainties and enhances energy resolution. There is a big difference between making things work in simulation and making things work in data to the degree that we trust the results,
Drielsma said.
ICARUS Benefits From SPINE
SPINE’s work helped ICARUS produce its first neutrino physics result in the search for sterile neutrinos. The tool recovered roughly 60% more usable neutrino interactions than conventional reconstruction, improving sample purity and efficiency. It also delivered a cleaner sample, improved energy resolution, and reduced systematic uncertainties.
ICARUS, which holds 760 tons of liquid argon, faces challenges from overlapping interactions and cosmic ray interference. SPINE’s ability to disentangle these signals has been critical. Without a way to disentangle them, we’d be stuck,
said Drielsma, highlighting the tool’s role in advancing neutrino research.
DUNE’s SPINE Integration
SPINE will be deployed on detectors, including DUNE, which is expected to start taking data in 2030. DUNE’s near detector will observe hundreds of overlapping neutrino interactions at once, creating an enormous reconstruction challenge. SPINE’s design makes it uniquely suited to handle this complexity, with models already trained for both DUNE’s near and far detectors.
Researchers hope SPINE will help answer fundamental questions. Until a year ago, we hoped this approach would work,
Drielsma said. Today, we believe it will. We’ve built the tool. Now we get to see what it discovers.
SPINE’s Creators
SPINE grew out of work by SLAC scientist Kazuhiro Terao, who began applying emerging computer-vision techniques to neutrino detector images in 2015 and 2016 as a postdoctoral researcher and recipient of a DOE Early Career Award. The goal was to record the evidence and then use that evidence to make the final conclusion,
Terao said.
SLAC physicist Mark Convery emphasized SPINE’s role in tackling smallish effects
with big physics impacts.
You need a really intense beam, a really capable detector and a really efficient and precise reconstruction technique,
he said. SPINE is that third leg of the stool.
SPINE’s Next Phase
As SPINE continues to evolve, its impact on neutrino physics—and the broader quest to understand the universe—will only grow.