Sound Analysis in Healthcare

What is possible and why it is important

Sound contains a huge amount of medical information just like visual information. It is easy to notice when someone is ill by just listening to their voice, also cough and sneeze are obvious clues of a cold.

There have been some academic researches on using sound analysis for medical diagnosis and screening. Many papers illustrated that there is clear evidence that exists in the audio although usually, it is under controlled, specific conditions. Finding evidence in the laboratory condition is important, but still, it was hard to be applied to daily real-world monitoring because such symptoms are minor, ambiguous, also vary a lot depending on the individual people.

However, recent advances in machine learning technology rapidly transform the healthcare industry. Now we start to see more smart healthcare services through mobile and wearable devices that are designed to be working in the real world. And the sound is something that is underinvestigated yet but many exciting services and applications are upcoming.

So, what kind of clues are hidden in the sound, and what is possible to identify through sound analysis?

Speech

Speech is often the easiest clue for noticing that someone feels unwell. It can be related to more physical change, but also mental health can influence speech sound. These are some diseases that can be relatively clearly identified by speech analysis.

• Parkinson’s disease (PD)
• Alzheimer’s disease (AD)
• Depression
• Cold, influenza, bronchitis, and pneumonia

Usually, PD affects voice to be quieter and more monotone pitch, also sounds more breathy and hoarse. In the case of AD, speech conveys a clue of mild cognitive impairment. For depression, people tend to have less variation in tones regardless of sentence. The absolute tone itself isn’t much important, but usually, depression makes people convey less emotion which results in a flatter mood. It is also obvious that people’s voices become hoarse when influenced by cold, influenza, bronchitis, and pneumonia which causes laryngitis.

Cough and sneeze

Cough analysis became one of the most important topics in the healthcare industry due to a recent COVID-19 pandemic. Institutions such as MIT and the University of Cambridge published studies about distinguishing a cough sound of COVID-19 patients. Surprisingly, it showed very high accuracy even for asymptomatic positive cases.

Although all the laboratory experiment has some limitation and also further research should be carried out to see whether it can be applied to all cough sounds from the patients who suffer from other kinds of diseases, it showed a great potential to use cough sound as a one of the quick, cheap, and fast prescreening methods for COVID-19.

In addition, the frequency of cough and sneeze sound is highly valuable information for monitoring patient’s status, also to analyse the risk of infection level in the public spaces.

Snore

The snoring sound might not sound very critical for healthcare, it does affect the quality of sleep. Accurate monitoring of snoring can be used for sleep analysis, also to measure how severe it is. There are increasing numbers of a mobile phone app that uses snoring sound analysis for sleep monitoring, although its performance is somewhat not good enough. In more serious cases such as apnea and hypopnea, the snoring analysis also can be the most obvious clue of its severity.

Digital stethoscope

Body-generated sounds contain a tremendous amount of information and this is the reason why doctors using a stethoscope as one of the basic tools for diagnosis. Recently, a digital stethoscope is getting more and more popular and it can be used for automatic analysis of body-generated sound. According to various researches in this field, Cardiopulmonary dysfunction (respiratory tones), cardiovascular abnormalities (heart rate analysis), and digestive abnormalities (small intestine, large intestine) can be analysed by the sound recorded by digital stethoscope.

Elderly care

Sounds generated from the human body are important, but other kinds of sound events are also closely related to healthcare services. More and more countries have an increasing proportion of the elderly population, and it becomes a problem in many countries. The sound analysis provides a less invasive way to monitor a home, possible to detect someone falls on the ground due to a slippery surface in a toilet, also screaming and shouting for asking for help. Of course, all other sounds we mentioned above such as snore, cough, sneeze, speech, and sounds obtained from digital stethoscopes can be integrated for elderly care services.

Conclusion

Measuring important things like heartbeat and blood pressure becoming more and more popular also advances in computer vision technology make our daily life much more convenient because it allows us to monitor various symptoms without physically visiting the hospital. Sound analysis is still not so widely adopted yet, but it has great potential to be the next big thing in the digital healthcare market. A microphone is already installed in many devices such as smart speakers and mobile phones, so sound-based healthcare can be helpful to a large number of people without buying new smart devices.

At Cochl, we provide an off-the-shelf technology for detection of various healthcare-related sounds such as cough, sneeze, snore, and scream, also expanding its ability to understand more detailed health-related information to make a safer and healthier world.