Researchers used new analysis techniques to independently validate the first photograph of a black hole at the center of galaxy M87, captured by the Event Horizon Telescope Collaboration in April 2019.

Samuel Lai and Nithyanandan Thyagarajan published their findings in Nature. They developed a new method called "closure invariants" that combines measurements from three telescopes at once instead of the standard practice of pairing two measurements. This approach eliminates atmospheric and instrumental errors that can affect radio telescope data.

The team also created a machine learning imaging method called GenDIReCT to reconstruct the black hole's appearance from the data. To minimize bias, they trained the model on ordinary images of cats, dogs, and cars rather than black hole images. They validated the method on real data from other galaxies before applying it to M87 observations on December 4, 2025.

Their independent results confirmed the iconic ring structure from the original EHT publication. The black hole's mass appeared consistent at 6.5 billion suns, and the ring diameter and orientation matched the original findings. Some differences in fine details like ring width and surface brightness appeared, but the overall structure was consistent.

Lai and Thyagarajan also developed a separate imaging method based on traditional statistics that validated their findings using the new closure invariant data. This resolved previous scientific disagreements about whether details in the original black hole image were accurate.