AI-Powered Algorithm to Revolutionize Detection of Atrial Fibrillation

This is because the irregular heartbeat in AFib can lead to the formation of blood clots in the heart, which can then trigger strokes. Additionally, AFib is known to raise the likelihood of heart failure or even death. To mitigate these risks in patients diagnosed with AFib, healthcare professionals typically prescribe anticoagulants, which are medications that help prevent blood clots. They may also recommend various lifestyle and medical therapies. However, diagnosing AFib can be challenging, as many individuals experience only sporadic episodes of irregular heartbeat or exhibit minimal symptoms. Common symptoms of AFib include heart palpitations, lightheadedness, shortness of breath, and chest pain. In cases where these symptoms are present, cardiologists generally conduct a detailed electrocardiogram (ECG) in their office, using ten electrodes placed on the patient's body to record heart rhythms for about ten seconds. If no immediate irregularities are observed, ongoing monitoring at home is usually recommended for a duration of one to two weeks, employing a simpler, wearable ECG patch that features a single electrode. However, this method may not always detect AFib in those who experience very infrequent episodes.

At Scripps Research (La Jolla, CA, USA), scientists have developed a novel artificial intelligence (AI) model aimed at enhancing the screening process for AFib. This model is adept at identifying subtle variations in normal heart rhythms that indicate a risk for AFib, which conventional screening methods might miss. The research utilized data from nearly half a million individuals, each of whom had worn an ECG patch for two weeks to record their heart rhythms. The AI model analyzed this data, identifying patterns that differentiated individuals with AFib from those without. This new model shows promise in improving the detection of those at risk for AFib, thereby potentially averting serious complications such as stroke and heart failure.

Notably, the machine learning model was more accurate at predicting AFib risk even after the researchers integrated all known AFib risk factors into their own, manual models—including demographic data and ECG measures such as the variability between different heart beats. It remained effective in predicting AFib risk across different age groups, including both older individuals, who have a higher risk of AFib, and those under 55, who typically have a lower risk and are often not included in general AFib screenings. While this model is not designed to diagnose AFib directly, it represents a significant step towards creating a screening test for those at heightened risk for AFib or exhibiting symptoms. This could mean wearing an ECG patch for just one day to ascertain if extended testing is necessary. Alternatively, the model might analyze data from one- or two-week ECG recordings to identify patients who should undergo repeat testing, even if no AFib is detected during the initial timeframe. The research team is planning a prospective study and aims to refine their models further by integrating additional data sources, such as electronic medical records.

“With this new tool, we can better identify patients at high risk of AFib for further tests and interventions,” says senior author Giorgio Quer, PhD, director of artificial intelligence at Scripps Research Translational Institute and an assistant professor of digital medicine at Scripps Research. “Long term, this can help drive the right resources to the right people and potentially reduce the incidence of stroke and heart failure.”

Related Links:
Scripps Research