The sound of a functioning heart, vessels, and lungs contains a surprising amount of information about their structure and function. These sounds are often sufficient to make an accurate diagnosis and conduct informative clinical monitoring. But there has always been a problem. Not surprisingly, mastering the ability to analyze these sounds has traditionally been a complex, long-term process. Add to that, the problem of so called low-frequency sounds. Low-frequency sounds are thousands of time harder for humans to hear so our consciousness leans towards ignoring them. Some think this is evolutionary. Most things in nature that could harm humans (angry bears, volcanoes, other humans etc.) tended to make sounds at much higher frequencies so we are predisposed to focusing on those. Yet a number of important auscultative symptoms, such as gallop tones S3 and S4, mitral stenosis murmur, Still’s murmur and Austin Flint murmur are exclusively low-frequency and extremely hard to perceive. With the help of modern analysis, sound processing and artificial intelligence technologies, this traditional limitation no longer exists and so neither do the barriers to using biological sounds to accurately, quickly and inexpensively diagnose conditions.
Modern echocardiography produces high quality medical insights but is not a panacea. Like all measurements, it has limitations based on the fundamental nature of the information it is based on; namely the shape and related mechanical movement of structures. To be clear it is a very rich information source from which to mine insights. But from a technical perspective, biological sound data is arguably just as rich if not more potentially informative. All moving (dynamic) systems emit energy as sound. What makes sound so rich is that it has two fundamental components: (1) Frequency: e.g. low-frequency oscillations, as a component of hemodynamic murmurs, are caused by movement of large blood volumes, and the high-frequency oscillations by high-speed blood flows etc. and (2) Amplitude. Amplitude considered alongside frequency provides additional information about the power of the sound. Power is interesting because we can use it to assess the severity of certain conditions (like mitral regurgitation). Diagnosis based on ultrasound or other tests can actually miss this because the underlying signals they process lack this information.
We think chest sound data is extremely valuable and, as it happens, potentially easier to collect and analyze than echocardiography for instance. And early accurate diagnosis matters. Diseases such as infectious endocarditis can be detected by the presence of quiet cardiac noise at an early stage but in practice it is often detected only much later with echocardiography, when the disease has caused catastrophic damage and the chances of survival are diminished.