# This AI Reads an ECG in Two Seconds and Beats Cardiologists at Diagnosis > Source: > Published: 2026-09-01 02:19:49+00:00 *Imperial's AI ECG tool is not trying to replace a cardiologist. It is trying to catch the heart failure and valve disease a routine ECG leaves hiding in plain sight.* Researchers at Imperial College London's National Heart and Lung Institute have put a sharp claim in front of cardiology's biggest European audience: an AI model can read a standard ECG in under two seconds and flag heart disease that a trained human cannot see on the trace. According to The Guardian, the system was presented at the European Society of Cardiology Congress in Munich, held from August 28 to 31, 2026, after testing across 67,000 patients in the United States. It identified up to 81% of heart failure cases and 90% of heart valve disease cases. That is the story. Not a shiny new scan. Not a hospital machine that only rich health systems can buy. The ECG has been sitting in clinics for more than a century, printing out the same electrical squiggles while doctors use it mainly to check rhythm, rate and obvious heart trouble. Now Imperial's team is saying there is more signal inside that cheap ten-second test than a person can read. ## The old test was not finished ESC Congress 2026 gave artificial intelligence its official spotlight this year, and the programme pointed directly at ECG interpretation, imaging analysis and risk prediction in routine care. More than 30,000 cardiology professionals were expected in Munich. The society also released its 2026 heart failure guidelines at the meeting, so El-Medany's work landed in the middle of a conversation about earlier recognition, not off to the side as a tech demo. Dr Ahmed El-Medany, a British Heart Foundation clinical research fellow who led the Imperial analysis, described the tool as 'superhuman AI.' That sounds grand. Then you get to the practical meaning. The model is not reading the ECG the way a consultant reads it. It is finding statistical patterns in the electrical trace that are invisible to the human eye. [Instagram Will Bury AI Influencers That Skip Its New AI Label](https://startupfortune.com/instagram-will-bury-ai-influencers-that-skip-its-new-ai-label/) Instagram has renamed its 'AI creator' label to 'AI-generated profile' and will cut off Reels and Explore recommendations for any account featuring an undisclosed AI-generated persona. The move follows reporting from Wired and The New York Times on fake AI influencers and doctors, and lands amid Meta's $18 billion settlement with U.S. states over... - [how to label AI generated profiles on Instagram](https://startupfortune.com/instagram-will-bury-ai-influencers-that-skip-its-new-ai-label/) - [Instagram AI influencer account reach penalties explained](https://startupfortune.com/instagram-will-bury-ai-influencers-that-skip-its-new-ai-label/) Here's the thing: speed is not the most interesting part. Two seconds makes a good headline, but the harder point is that the software is pulling a different kind of information from a test doctors already order every day. A faster cardiologist is useful. A test that sees a hidden disease signal is more serious. ## Cardiovolt.ai is the commercial bet The research comes from Professor Fu Siong Ng's group at Imperial, with long-running support from the British Heart Foundation. Now it's a company. The group is moving into a spinout called Cardiovolt.ai, which Imperial said raised £1.4 million in pre-seed funding led by Twin Path Ventures. The funding also includes support from Imperial's DT Prime fund, Innovate UK and Imperial Enterprise. The founding team matters here. This is regulated medical software, not a wellness app with a nicer dashboard. Imperial and BHF name Professor Ng as chief medical officer, Dr Libor Pastika as chief technology officer, Dr Arunashis Sau as chief scientific officer and Boroumand Zeidaabadi as chief executive. Companies House records also list Ng, Pastika, Sau and Zeidaabadi Nezhad as officers of Cardiovolt.ai. Imperial's published figures put the broader model family at 83% to 93% diagnostic accuracy for heart disease. For non-cardiovascular conditions such as diabetes and kidney disease, the range drops to 70% to 80%. All from a single ten-second ECG. A separate 2025 Imperial-led study in the European Heart Journal found AI-enhanced ECGs could identify future risk of leaky heart valve disease in about 69% to 79% of cases. Patients flagged as high-risk were up to ten times more likely to develop the disease than those marked lower risk. Those numbers are strong. They are not magic. ## The hospital arithmetic is blunt An ECG is cheap and quick. An echocardiogram, the scan doctors use to confirm valve disease and assess heart structure, needs equipment, trained staff and appointment capacity. If you run the AI over routine ECGs and send only the suspicious cases to imaging faster, you change the queue. You do not need every patient to become a cardiology case. You need the right patients to stop waiting unnoticed. The limits are just as plain. The tool cannot diagnose heart failure or valve disease on its own, and the researchers are not selling it as a replacement for an echocardiogram or a cardiologist's judgment. A false positive means a patient may be sent for a scan they did not need. A false negative is worse, because the disease stays hidden. At 81% for heart failure and 90% for valve disease, some cases still get missed. [Pentagon Puts ChatGPT and Grok Side by Side for 3 Million Personnel](https://startupfortune.com/pentagon-puts-chatgpt-and-grok-side-by-side-for-3-million-personnel/) On August 31, 2026, the Department of War added OpenAI's ChatGPT Mil and Starshield AI's Grok for Government to GenAI.mil, its secure military AI portal already used by 1.7 million of 3 million eligible personnel. The Pentagon says pairing the two chatbots is a deliberate hedge against vendor lock-in, even as neither OpenAI nor xAI has publicized... - [military AI tools ChatGPT and Grok comparison](https://startupfortune.com/pentagon-puts-chatgpt-and-grok-side-by-side-for-3-million-personnel/) - [Pentagon deploys ChatGPT Grok to three million personnel](https://startupfortune.com/pentagon-puts-chatgpt-and-grok-side-by-side-for-3-million-personnel/) Still, you can see why hospitals will pay attention. One ECG, no extra hardware, an answer in under two seconds, and a triage signal that a human reader cannot pull from the paper unaided. That is not enough to declare victory. It is enough to make the old ECG look unfinished. **Also read:** [Instagram Will Bury AI Influencers That Skip Its New AI Label](https://startupfortune.com/instagram-will-bury-ai-influencers-that-skip-its-new-ai-label/) • [Banks Financing Asia's AI Data Center Boom Are Hitting Their Own Lending Limits](https://startupfortune.com/banks-financing-asias-ai-data-center-boom-are-hitting-their-own-lending-limits/) • [Pentagon Puts ChatGPT and Grok Side by Side for 3 Million Personnel](https://startupfortune.com/pentagon-puts-chatgpt-and-grok-side-by-side-for-3-million-personnel/)