Two Master of Science students at the Technion developed a machine learning algorithm capable of accurately predicting whether a patient will develop atrial fibrillation within the next five years.
The researchers intended to determine whether a machine learning algorithm could detect characteristics suggestive of atrial fibrillation even though no human cardiologist had diagnosed atrial fibrillation at the time.
Atrial fibrillation is an irregular cardiac rhythm that does not cause immediate mortality but dramatically raises the risk of stroke and death in patients.
Will you offer us a hand? Every gift, regardless of size, fuels our future.
Your critical contribution enables us to maintain our independence from shareholders or wealthy owners, allowing us to keep up reporting without bias. It means we can continue to make Jewish Business News available to everyone.
You can support us for as little as $1 via PayPal at office@jewishbusinessnews.com.
Thank you.
Notifying patients that they are at risk of getting it gives them time to make lifestyle changes that will help them avoid or delay the start of the condition.
Additionally, it may stimulate regular follow-ups with the patient’s cardiologist, ensuring that if and when the issue arises, it is swiftly detected and treated. Sedentary lifestyle, obesity, smoking, and genetic susceptibility are all known risk factors for atrial fibrillation.
Shany Biton and Sheina Gendelman, working under the supervision of Assistant Professor Joachim A. Behar, head of the Artificial Intelligence in Medicine laboratory, trained a deep neural network to recognize patients at risk of developing atrial fibrillation within five years using over one million 12-lead ECG recordings from over 400,000 patients.
Then, they merged the deep neural network with clinical data about the patient, including some of the known risk factors such as ECG recordings and the patients’ electronic health records provided by the Telehealth Network of Minas Gerais (TNMG). This public telehealth system assisting 811 of the 853 municipalities in the state of Minas Gerais, Brazil.
The resulting machine learning model successfully predicted the development of atrial fibrillation risk in 60% of cases while maintaining a high specificity of 95%, implying that just 5% of those identified as possibly at risk never developed the condition.
“We do not seek to replace the human doctor – we don’t think that would be desirable,” said Prof. Behar of the results, “but we wish to put better decision support tools into the doctors’ hands. Computers are better equipped to process some forms of data. For example, examining an ECG recording today, a cardiologist would be looking for specific features which are known to be associated with a particular disease. Our model, on the other hand, can look for and identify patterns on its own, including patterns that might not be intelligible to the human eye.”
Due to the inexpensive cost of ECGs as routine tests, the machine learning model might readily be integrated into clinical practice, hence improving healthcare management for a large number of people. Access to additional patient datasets would allow the system to improve incrementally as a risk prediction tool. Additionally, the algorithm could be extended to predict additional cardiovascular diseases.
