Greek AI Tool Diagnoses Heart Issues With 90% Accuracy

A new AI tool developed by Greek researchers is catching deadly heart problems that doctors often miss, and it works with equipment already in hospitals worldwide.

Interventional cardiologist Polydoros Kampaktsis and data scientist Archontis Giannakidis created the software at Greece's Archimedes AI Research Centre. Their innovation solves a critical problem: the right side of the heart has such a complex shape that standard ultrasounds often can't reliably detect when something's wrong.

That's a serious issue. Right ventricle dysfunction shortens life expectancy even when other heart factors look healthy, and up to 20% of patients getting echocardiograms show abnormalities in this area. Until now, the only way to accurately assess right ventricle function was through cardiac MRI, which costs significantly more and isn't widely available.

The AI software transforms regular two-dimensional ultrasound images into highly accurate diagnoses. A cardiologist takes standard measurements, then the algorithm analyzes complex patterns hidden in the data that human eyes can't detect. The system runs on minimal computing power, meaning it works on laptops and even mobile phones.

Initial testing at Columbia University Medical Center in New York involved 50 patients and showed accuracy approaching that of cardiac MRI. Researchers then launched a larger trial at Athens' Onassis Cardiac Surgery Center with 200 patients who received both ultrasounds and MRIs for comparison.

The results from the first 120 patients exceeded expectations: over 90% accuracy. Cardiac surgeon Dimitris Avgerinos called it a potential "real revolution" for everyday heart care.

The Ripple Effect

The technology arrives at a crucial moment for global healthcare. Millions of people worldwide lack access to expensive cardiac MRI machines, particularly in rural areas and developing countries. This AI tool democratizes advanced heart diagnostics by working with ultrasound equipment that most clinics already own.

The researchers aren't stopping at the right ventricle. They're planning a multicentre study across hospitals in Greece and the United States to test how the technology performs in routine clinical practice. They're also working on full automation and adding uncertainty metrics so doctors know exactly how confident they can be in each result.

Similar AI tools for assessing the right ventricle using standard ultrasound simply don't exist yet. The breakthrough became possible only through recent advances in deep machine learning, technology that didn't exist just a few years ago.

The team emphasizes careful validation over hype. Kampaktsis stressed that AI medical applications should be evaluated based on real clinical benefit, not inflated expectations. Giannakidis pointed to important challenges ahead, including protecting patient data, complying with privacy regulations, and ensuring transparency in how algorithms make decisions.

Within months, more hospitals could start using this tool to catch heart problems before they become fatal.