AI reveals hidden brushwork in Renaissance paintings, opening new path to art authentication
Case Western Reserve University researchers have developed a method to analyze Renaissance paintings at the scale of a single hair, using artificial intelligence to map microscopic surface textures and uncover evidence about authorship and artistic technique.
The team scanned two paintings by El Greco and created ultra-detailed topographic maps capturing tiny ridges and grooves left by brushstrokes. They then trained an AI system to detect patterns across centimeter-scale areas that are invisible to the human eye.
The approach treated each painting as a network of interconnected pieces. The algorithm could determine whether surface patterns pointed to a single artist's hand or multiple contributors.
Finding unity in disputed works
One painting, "Christ on the Cross" at the Cleveland Museum of Art, showed striking surface uniformity. The second work, "Baptism of Christ" in Toledo, Spain, revealed something more significant: evidence of a single hand where scholars had long believed the master's workshop finished the painting posthumously.
The findings suggest a single set of materials, or possibly a single artist, created regions previously attributed to different hands. If confirmed, this could reshape scholarly understanding of El Greco's late work.
Michael Hinczewski, associate professor of physics at Case Western Reserve, said the technique offers a data-driven method to answer long-standing attribution questions. "When you can analyze details down to the width of a single paintbrush bristle, you start to uncover a kind of fingerprint," he said.
From casual conversation to seven-year collaboration
The project began unexpectedly: two Case Western Reserve graduate students who were dating-one studying art history, the other physics-started a conversation that became a collaboration. Seven years later, the work now involves scientists, art historians, and partners at the Cleveland Museum of Art, Cleveland Institute of Art, and the Factum Foundation in Madrid.
Andrew Van Horn, now a postdoctoral fellow at Purdue University, worked on the research at Case Western Reserve. He said applying computational methods to actual art history questions created both a new AI method and a discovery about El Greco. "Interdisciplinarity is going to be a key driver of innovation," he said.
Scaling up for authentication and forgery detection
Hinczewski envisions applying the technique across larger collections. Comparing surface "fingerprints" from different works could confidently attribute paintings, track how an artist's style evolved, and resolve disputes over disputed pieces.
As the database grows, the approach could also detect subtle inconsistencies pointing to modern imitations. Museums and collectors would gain a tool for identifying counterfeits.
The research appears in Science Advances. Hinczewski said the work is preliminary. "Even a few millimeters of paint can carry a wealth of information about how a work was made. As these tools evolve, they could transform how we study artists over time-and how cultural heritage is protected."
For researchers working at the intersection of science and art, this approach demonstrates how AI for Science & Research can solve domain-specific problems that traditional methods cannot address.
Your membership also unlocks:
