Heart sounds typically consist of two regular sounds, known as S1 and S2, for every heartbeat. The normal blood flow inside the heart is mainly laminar and therefore silent; but when the blood flow becomes turbulent it causes vibration of surrounding tissue and hence the blood flow is noisy and perceivable, causing an audible murmur. Murmurs are often indicative of pathological changes, but they overlap with the cardiac beat and cannot be easily separated by the human ear using a traditional stethoscope.
Eko's algorithms are designed to differentiate between healthy heart sounds and those containing heart murmurs using a deep neural network model trained on multiple datasets of heart sounds.
When analyzing a single lead ECG, Eko's algorithms can detect the presence of atrial fibrillation and normal sinus rhythm using state-of-the-art machine learning techniques. The algorithms use an ensemble of deep neural network models that have been trained and validated across multiple datasets to ensure they perform accurately and robustly on real-world data. Furthermore, the models are capable of detecting comorbidities including tachycardia, bradycardia, and premature ventricular contractions (PVCs).
Eko has also partnered with the Mayo Clinic to develop a first-of-its kind heart failure screening algorithm that uses DUO's single lead ECG to detect asymptomatic left ventricular dysfunction via detecting low Ejection Fraction.
By screening with Eko’s ECG algorithms, patients with undiagnosed rhythm abnormalities or heart failure could be detected earlier and start preventive therapy earlier meaning future adverse events, including strokes, might be avoided.