Particle.news
Download on the App Store

AI Sifts 35 Years of Hubble Images, Flagging Over 1,300 Cosmic Anomalies

The peer-reviewed scan demonstrates AI-driven anomaly detection as a practical triage for upcoming survey-scale datasets.

Overview

  • Researchers used a custom system called AnomalyMatch to scan nearly 100 million Hubble image cutouts in about two and a half days.
  • Expert review confirmed roughly 1,300–1,400 true anomalies, with more than 800 not previously documented in the scientific literature.
  • Flagged objects include merging and interacting galaxies, gravitational-lens candidates, jellyfish galaxies, irregular star-forming regions, and edge-on disks.
  • The work, led by ESA-affiliated scientists David O’Ryan and Pablo Gómez, is detailed in Astronomy & Astrophysics and summarized by NASA/ESA.
  • The catalog provides prioritized targets for human follow-up and a template for processing far larger datasets from Euclid, the Rubin Observatory, and NASA’s Roman Telescope.