Researchers at University College London have developed an AI system capable of detecting Parkinson's disease up to seven years before patients show any clinical symptoms.
The tool, described in a study published in Nature Communications, analyzes eight simple biomarkers found in routine blood tests. By identifying subtle patterns invisible to the human eye, the AI can flag individuals at high risk with remarkable accuracy — correctly identifying 96 percent of cases in clinical trials.
"This is a game-changer for early intervention," said Dr. Kevin Mills, who led the research team. "Parkinson's is currently diagnosed based on motor symptoms like tremors, but by that point, patients have already lost 50 to 70 percent of their dopamine-producing neurons."
The eight biomarkers are proteins already measured in standard blood panels, meaning the screening could be integrated into routine healthcare without expensive new equipment. Parkinson's affects over 10 million people worldwide, with numbers expected to double by 2040.
Early detection has been the holy grail of Parkinson's research for decades. While several drugs have shown promise in slowing neurodegeneration, they have failed in trials — not because they don't work, scientists believe, but because they are administered too late.
"Imagine being able to start neuroprotective therapy seven years before a single tremor," said Dr. Cristina Simonet, a co-author. "We're not just diagnosing earlier. We're creating the possibility of prevention."
The team validated their model using data from the UK Biobank, containing health records from over 500,000 participants. Clinical trials are now being planned to test whether early intervention guided by AI screening can meaningfully alter disease outcomes.
For the millions of families affected by Parkinson's, the message is one of hope: the era of waiting for symptoms may soon be over.
