Poland-based medical technology company Medicalgorithmics announced that its novel AI algorithm, DeepRhythmAI (DRAI), is superior to human specialists in detecting heart disorders.
The company’s announcement is based on the results from a recent study of its DRAI, dubbed DRAI MARTINI, presented at the ESC Congress 2024.
In the study, DRAI technology has shown significantly higher sensitivity than human specialists in detecting heart rhythm disorders on long ECG recordings.
The study found that traditional methods of reading ECG using human technicians missed diagnoses 14 times more often than when using DRAI technology.
Medicalgorithmics CTO Przemysław Tadla said: “The DRAI MARTINI study underscores the value of AI in augmenting the capabilities of ECG technicians and cardiologists.
“We anticipate that AI will become a standard tool in cardiac diagnostics, supported by research such as the DRAI MARTINI study.”
The DRAI MARTINI study evaluated whether the AI algorithm could replace human technicians for reporting ECG data, using data from more than 14,600 patients.
In the study, the patients were monitored on average for 14 days for symptoms such as palpitations, fainting, and dizziness.
DRAI’s performance was compared with traditional ECG analysis conducted by technicians, in collaboration with 50 top cardiologists from the US and Europe.
The AI technology significantly reduced false negative findings compared to traditional methods.
In the study, false negative findings occurred in only 3.2 per 1,000 patients using AI, compared to 44.3 per 1,000 patients for human technicians.
The results showed a superior sensitivity of 98.6% for DRAI in detecting critical arrhythmias, compared to 80.3% sensitivity for ECG technicians.
The high sensitivity of AI technology turned into a negative predictive value of 99.9%, which ensured the accurate identification of almost all patients without arrhythmias.
Lund University associate professor and the study lead researcher Linda Johnson said: “Long-term ECG analysis is time-consuming and when human technicians read ECG there is potential for error, leading to missed diagnoses of clinically meaningful arrhythmias.
“Our study tests a state-of-the-art AI algorithm and finds that the AI had a 14 times lower risk of missing a critical arrhythmia diagnosis.”
