Cardia AI is a Czech medtech startup that develops long-term ECG monitoring with detection of cardiac arrhythmias using artificial intelligence. The talk showed why early detection of arrhythmias is so challenging, what the limits of routine practice are, and how long-term measurement can change the situation. It is not a miracle solution, but a practical tool that can increase the chances of identifying high-risk conditions.
Why early arrhythmia detection is a problem
A large part of the population has heart problems, but approximately 40 % of those affected have no symptoms and therefore are unaware of their condition. The population is aging, and in the European Union alone around 200,000 new cases of atrial fibrillation, one of the most serious arrhythmias, are added each year. In Slovakia, heart disease is associated with roughly 45 % of deaths, and annual treatment costs reach billions. The available preventive care, however, is not always sufficient to catch episodic, "elusive" arrhythmias.
Limits of routine testing and what it implies
A brief, 30-second ECG at a general practitioner detects an arrhythmia in about 1 % of cases, since episodes occur randomly and in clusters. The gold standard, a 24-hour Holter monitor, increases the chance of detection to about 5 %, but the arrhythmia can still slip through. In the Czech Republic, the average wait to see a cardiologist is 60 to 83 days, so patients often spend months in a cycle of tests without a clear result. Untreated and late-detected arrhythmias can lead to heart failure, stroke, or an earlier onset of dementia.
Long-term measurement with AI: Cardia AI's approach
Cardia AI uses a sports chest strap and a mobile app so that a person can measure anywhere and for as long as needed. After the measurement, the data are sent to the cloud, where AI removes artifacts and noise and analyzes the clean recording to detect arrhythmias; the result appears in the app within about five minutes. For atrial fibrillation, the solution achieved a detection rate of around 23 %, which is roughly five times that of a 24-hour Holter monitor in the given comparison. The user sees a simple "traffic light": green means no findings, yellow signals a non-serious finding, and dark orange flags severity with a recommendation to see a doctor as soon as possible.
