Artificial Intelligence analyzes up to 100% heart failure from heart sounds

Unwrapping aortic valve dysfunction through complex network analysis: A biophysics approach developed by a research team from the University of Kerala, India and the University of Nova Gorica, Slovenia, reveals that aortic valve stenosis causes heart failure using a digital stethoscope and a machine learning model. It is a sound imaging technology that can detect diseases.

 Since the test for aortic stenosis, which is a conventional method that requires high costs, can be performed at a low cost, we hope that in the future it will be made available through a mobile app and used in areas where advanced medical care is not available. It is said that there is
Artificial Intelligence analyzes up to 100% heart failure from heart sounds

 Aortic stenosis is one of the four valves in the heart (the valve that separates the left ventricle and the aorta), and the aortic valve (the valve that separates the left ventricle and the aorta) does not move well, and the valve opening narrows, allowing blood to flow throughout the body. refers to a state in which it becomes difficult to In severe cases, it can lead to heart failure.

 To make matters worse, mild aortic stenosis is difficult to cause symptoms, and because it progresses gradually, there are long periods of no symptoms. In addition, since the diagnosis of aortic stenosis requires advanced technology, it is said to be difficult to diagnose at hospitals and clinics that do not have advanced technology, and early detection is difficult.

 In our research, we developed a method to identify valve dysfunction by analyzing heart sound data acquired with a digital stethoscope using a machine learning model.

 The sound of the heart is when the mitral and tricuspid valves out of the four valves close, making a “lub” sound, and the ventricle relaxes. It pauses when it fills with blood and makes a second “dub” sound as the aortic and pulmonary valves close. This repetition produces a continuous beat.

 However, in patients with aortic stenosis, the sounds “lub” and “dub” can be heard at the same time. Since it is different from a healthy heart, the difference is identified to determine aortic stenosis or normal.

 The sound is recorded for about 10 minutes, and the measured sound data is converted into a dot graph before analysis. A healthy heart displays two distinct clusters of points on the graph, while a heart with aortic stenosis results in a poorly defined scattered point cloud. Machine learning is used to analyze this difference and classify those with and without the disease.

 As a result of evaluation experiments, it is said that a maximum classification accuracy of 100% was achieved. Unlike other methods that consider only signal strength, it is thought that the accuracy was high because it took into account the correlation of each point.

 The research team believes that this sound imaging technology can be applied not only to the heart but also to diagnose other organs