שיינה גנדלמן ושני ביטון
שיינה גנדלמן ושני ביטוןצילום: הטכניון
Graduate students from Israel's Technion in Haifa have developed an advanced 'deep learning' computer algorithm that can calculate the chances of developing atrial fibrillation in the next five years - a heart disorder that can lead to critical results.

Shani Bitton and Sheina Gendelman, graduate students at the Technion's Biomedical Engineering Department developed the 'deep learning' algorithm, under the direction of Dr. Joachim Behar, head of the Artificial Intelligence in Medicine Laboratory(AIMLab) at the Technion.

Atrial fibrillation is a heart rhythm disorder that is not immediately life-threatening but significantly increases the risk of stroke and death.

It is now known that certain behaviors such as sedentary lifestyle, smoking and obesity increase the risk of atrial fibrillation,so warning of such a risk may allow a person to take risk-reducing measures and enter a follow-up routine that will allow early detection of the problem.

The students trained a deep learning system (layered neural network) using more than a million ECG records of more than 400,000 patients, thus creating a mechanism to predict human chances of developing atrial fibrillation over a five-year period. They then combined the deep neural network with clinical information on the patient.

This model was able to correctly predict the risk of developing atrial fibrillation in 60% of cases,
while maintaining a high specificity rate of 95% (i.e. only 5% of the people identified as people at risk did not develop the disease).

Dr. Behar said: “We do not intend to replace the human doctor, but we want to give him better tools to support him in making decisions. Advanced computational tools know how to process data more efficiently and accurately than any human being, and deep learning enables the identification of patterns that were unknown to us.

"Throughout history, physicians have progressed from manual pulse measurement to a stethoscope and from there to an ECG,and we believe that ECG surgery based on machine learning is another important step that has significantly improved the quality of diagnosis and prevention."

According to the researchers, because ECG is a routine and relatively inexpensive test,
the machine learning model can be integrated into clinical practice and thus improve healthcare management.Access to additional data sets will allow the algorithm to gradually improve as a risk prediction tool for other cardiovascular diseases or complications.

The study was conducted in collaboration with Antonio Ribeiro of the University of Uppsala in Sweden and with Gabriela Miana, Carla Moreira and Antonio Luiz Ribeiro from Universidade Federal de Minas Gerais in Brazil. Patients' ECG records and electronic medical records were provided by the Telehealth Network of Minas Gerais,a public telemedicine system that assists most local authorities in the state of Minas Gerais, Brazil.