Stethoscopes, Electronics, and Artificial Intelligence

For all the advances in medical diagnostics made over the last two centuries of modern medicine, from the ability to peer deep inside the body with the help of superconducting magnets to harnessing the power of molecular biology, it seems strange that the enduring symbol of the medical profession is something as simple as the stethoscope. Hardly a medical examination goes by without the frigid kiss of a stethoscope against one’s chest, while we search the practitioner’s face for a telltale frown revealing something wrong from deep inside us.

The stethoscope has changed little since its invention and yet remains a valuable if problematic diagnostic tool. Efforts have been made to solve these problems over the years, but only with relatively recent advances in digital signal processing (DSP), microelectromechanical systems (MEMS), and artificial intelligence has any real progress been made. This leaves so-called smart stethoscopes poised to make a real difference in diagnostics, especially in the developing world and under austere or emergency situations.

The Art of Auscultation

Since its earliest appearance in 1816 as an impromptu paper cone rolled by Dr. René Laennec, partly to protect the modesty of his female patients but also to make it possible to listen to the heart sounds of an obese woman, designs for stethoscopes have come to the familiar consensus configuration: a chest piece, a pair of earpieces, and a tube or tubes connecting them. The chest piece consists of either a broad, thin diaphragm or an open-ended bell, while the earpieces are designed to fit snugly into the practitioner’s ears under light spring pressure to occlude as much environmental noise as possible.

State of the art – a Littmann Cardiology IV stethoscope. Source: 3M Littmann

The chest piece is pressed directly against the patient’s body and acts as an impedance matcher that couples faint vibrations from the watery interior to the air outside. The column of air in the tubing connected to the chest piece conducts vibrations up to the earpieces and onto the eardrums of the practitioner. Medically, the process is referred to as auscultation, and though the sounds thus heard are faint and often coupled with noise both from inside and outside the patient, with practice they can help paint a surprisingly complete diagnostic picture.

While auscultation is used for sounds generated all over the body, it’s mostly used for sounds made in the chest by the heart and by the lungs. The choice of which chest piece to use depends on which the frequency of the sounds that need to be heard. Lung sounds are generally of higher pitch, in the range of 200 to 2,000 Hertz, as air vibrates over and around the structures of the respiratory tract. The diaphragm chest piece is usually used for these sounds, with the thin, taut disc collecting sound over a wide area and coupling as much acoustic energy as possible up to the earpieces. Heart sounds are generally lower pitched, between 20 and 200 Hertz, and are caused by the turbulence of blood flowing through vessels and the mechanical sounds of the heart’s valves. Such sounds are best heard using the bell chest piece, which essentially uses the patient’s skin as a diaphragm. Many stethoscopes have dual-headed chest pieces that can be swiveled between diaphragm and bell.

No matter which chest piece is used, auscultation is more of an art than a science. While it’s an extremely sensitive instrument, a stethoscope is fairly broadband and couples every noise and vibration up the pipe and into the ear. A human body can be a noisy place, with bowel sounds (or borborygmi, perhaps the best word in the English language) interfering with lung sounds or the patient’s nervous chatter overriding a faint heart murmur. Add to that the tendency of a chest, which is essentially a big, hollow drum, to couple and resonate environmental noises, and picking a weak signal from broadband noise can be a challenging task.

Canceling the Noise

It seems like the stethoscope is a perfect target for electronic improvements — add a microphone to the chest piece, wire it to a small amplifier, and use earbuds instead of earpieces. Of course that’s been tried many times since the dawn of the transistor age and the small, powerful amplifiers it made possible.

So why haven’t doctors been sporting electronic stethoscopes for years? One reason is tradition: few tools are as strongly associated with a profession as the stethoscope, and to some degree, the instrument represents the trust that ideally exists between practitioner and patient. There are practical reasons for this apparent stasis, too. Chief among these is the stethoscope’s simplicity — no batteries required, fits in a pocket, and is ready to go as soon as the earpieces are placed. It doesn’t need to boot up, it doesn’t suffer from electrical interference, it has few parts to wear out or break down, and is easily cleaned and disinfected. Plus it’s cheap, a decent scope can be had for a couple of dollars.

Current version of the smart stethoscope. Source: Zebadiah Potler/Johns Hopkins University

But time and technology may finally have caught up with the old acoustic stethoscope, with perhaps the promise of improving patient care and saving lives. Johns Hopkins University recently announced a new “smart stethoscope” that completely replaces the acoustic system while maintaining the same form-factor as a traditional stethoscope. The smart stethoscope has an electronic chest piece using MEMS transducers to pick up body sounds, with sensitive, compact amplifiers driving small speakers in the earpieces.

If that were all the Johns Hopkins invention offered, it wouldn’t be much of an advancement. However, the Hopkins team also attacked the biggest problems with auscultation: noise. By incorporating some of the same technology that noise-canceling headphones employ, the smart stethoscope uses an additional microphone to pick up environmental noise and filter it out algorithmically. A research paper (PDF link) discusses the details; the video below shows a prototype at work filtering out the cries of children in a pediatric clinic coupling through the patient’s chest cavity. Notice too that the patient’s heart sounds are eliminated, allowing lung sounds to come through clearly.

The implications of an effective, affordable noise-canceling stethoscope can’t be overstated. Worldwide, very few medical interventions occur in a quiet setting. Crowded clinics, busy emergency rooms, active battlefields, and the backs of ambulances are where medicine is practiced, and where decisions are made based on what is heard in a stethoscope, or what is missed. Being able to filter out that noise and bring the signal to the front will make auscultation less art and more science, and hopefully improve outcomes.

What’s more, since the Hopkins smart stethoscope is a digital instrument, it can serve as input for an artificial intelligence application that automatically analyzes the data for signs of pathology. The system has been trained to identify the characteristic but oftentimes subtle sounds of pneumonia, a leading killer of children worldwide, using data from 1,500 patients. The app currently discriminates between healthy patients and those with pneumonia 87 percent of the time; apps for other diseases are currently being developed.

The Hopkins smart stethoscope has a lot of potential, especially if the cost per unit can be kept low. Most electronic stethoscopes currently run to $500 a piece or more, making access to them difficult in the developing world where they can arguably do the most good. It will be difficult to match the price point or simplicity of the tried and true acoustic instruments, but it seems like the next generation of electronic stethoscopes might just be a better deal in terms of outcomes.

19 thoughts on “Stethoscopes, Electronics, and Artificial Intelligence

    1. There are so many examples of where AI has arrived at a solution that is so far off the track it is not funny. AI does not have common sense. When talking medical diagnosis, a small mistake can lead to disasterous consequences (and lawsuits).

    2. It is more properly termed an expert system. Computers are very good, and fast, at analyzing things and finding signals in the noise once the algorithm has been developed. Doctors can be trained at it, but there is a limit to how much they can remember and discriminate. The best use is as a screening tool. Let the computer find what it can and tell the Doctor what to look more carefully at.

  1. As tech improves I suspect doctors will end up with more complex equipment maybe a handheld echo?

    Why try and hear turbulence when it’ll be pretty red and blue on the pulse Doppler?

    Why try and heart mitral regurg when you can see it in the flow dynamics of the valve

    Certainly that requires some training but what the heck else are we supposed to do once AI gets involved

    1. There is already a handheld ultrasound you can pair to a smartphone: the butterfly IQ. The last time I looked into it you needed to be a doctor/physician to buy one (says the FDA).

  2. “… it seems strange that the enduring symbol of the medical profession is something as simple as the stethoscope.”

    Not really. Simplicity, and the reliability, as well as low-cost that rides with it explain much. Especially if one’s in a cost-conscious, needs to work every time, can cobble from anything handy, environment. Any alternative will have a high-bar to jump.

  3. There are many stethoscope mods but the best I ever saw had nothing to do with improving performance and was sported by a senior ITU trainee in Denmark back in the 90’s. Their operating theatre “blues” were actually white and rather stylish. The trousers even had proper pockets and belt loops. A quick disconnect (borrowed from the cuff of an automatic blood pressure machine) had been fitted half way along the length of the stethoscope. This allowed the earpiece ends to be put into the left pocket and the metal end into the right. The tubing from each half was then threaded backwards through the belt loops on each side and joined at the back using the quick disconnect. The point of this complexity was to carry the stethoscope in a suitably “cool” manner and when needed, the pièce de résistance, he could could click open the disconnect in preparation, then reach down and whip out each half in Wild West twin Colt 45 gunslinger style, thus impressing (or not) the nurses on ITU. Happy days.

  4. An electronic stethoscope needing batteries is unlikely to improve the diagnosis of the common infectious diseases of childhood or rheumatic fever in the developing world. Improved detection of subtle renal artery stenosis, carotid artery stenosis, early pulmonary fibrosis or mitral regurgitation won’t really alter management in a resource constrained environment either.

    The money would probably be better spent on improving local food production, sanitation works and education of women to reduce malnourishment, epidemic diarrhoeal disease and improve outcomes for women, respectively.

      1. There is no such thing as a silent helicopter. :-)

        But from my point of view, improving healthcare in my part of the world, where I live (Europe) affects me more than things for a “resource constraint environment”. So this is not my primary criterion.
        On the other hand “medical equipment and pharma industries” do not lose much if cheap improvements for poor countries exist, that can not afford the expensive stuff anyway.

  5. Funny. So much discussion on ai intervening in diagnosis. I would have thought the potential was in being able to ‘remote’ monitor ie. remote doctor remotely guiding placement with high fidelity acoustics transmitted to his/her remote ears. Developing countries seem to have embraced cellular tech for obvious reasons but vehicle infrastructure sucks in many places. Doctor can’t bring the stethoscope to you? Then bring the stethoscope (remotely) to the doctor. As for ‘ai’ supplanting doctors: ‘ai’ is arguably better than doctors now, but you can’t sue them … so doctors will be safe until that is figured out :)

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.