We could improve our chances of spotting the early warning signs of heart disease and strokes if we could regularly check our blood pressure at home. Professor Dingchang Zheng may have found an innovative solution, by teaching computers to ‘listen’ to stethoscope sounds.
Medical science may be advancing rapidly, but some things are slow to change. To take a patient’s blood pressure 100 years ago, a doctor needed an arm cuff, a pressure gauge and a stethoscope. If that equipment sounds familiar, it is because it is exactly what GPs still use today.
‘Blood pressure’ is the strength with which blood pushes on your arteries as it flows, and the traditional approach to measuring it is pretty straightforward. The cuff is wrapped around the patient’s arm and tightened, restricting the blood flow. Then pressure is gradually released, until the blood flows normally.
Your doctor listens for two points: the sound that blood makes as it re-enters the artery (known as ‘systolic pressure’) and the moment when the sound disappears as the flow returns to normal (‘diastolic pressure’). These readings are used to assess whether your blood pressure is healthy, high, or low.
For Professor Dingchang Zheng, Professor of Medical Technology Innovation at our University, this venerable system remains the gold standard. “I would recommend it to anyone,” he adds. “It’s still the most accurate, non-invasive way to take blood pressure for clinical diagnosis.”
But this measurement also requires considerable training. Meanwhile, there is a growing argument that we should also be measuring blood pressure regularly at home.
"It’s becoming normal for the public to measure their weight at home. I think it’s more important to check blood pressure."
High blood pressure puts a huge strain on our arteries and organs, and increases the risk of serious conditions like coronary artery disease, congestive heart failure, and strokes. “It is becoming normal for the public to measure their weight at home,” Dingchang says. “I think it’s more important to check their blood pressure, because it’s a leading risk factor for so many diseases.”
Since not everyone can be trained to use a stethoscope to measure systolic and diastolic pressures, the ideal way to enable this would be through an automated home device that does it for us. But although many automatic devices exist, they are much less accurate than the manual approach, because they only estimate blood pressure using equations, rather than listening for pressure points as clinicians do.
Dingchang has spent his career trying to find a better solution. Through his work in our School of Allied Health, he may finally have an answer.
His approach relies on machine learning: programming computers to look for patterns in data and make decisions based on that information. “We are probably the only research group in the world using machine learning to improve automatic blood pressure measurements,” Dingchang says.
“Machine learning is transforming our research. We can now teach computers to listen like a human ear."
His group has used a database of stethoscope sounds recorded during blood pressure measurements, and depicted as a set of digital patterns. By zooming in on the image associated with the stethoscope sound of each heartbeat, it becomes possible to see that each sound has its own, distinctive features. These individual patterns, or ‘digital signatures’, can be used to train a computer to spot the systolic and diastolic pressure points. Effectively, the computer can be programmed to ‘listen’ for these, just like a clinician.
Dingchang’s team has developed a machine-learning algorithm that enables a computer to do exactly that. They also evaluated this using data from 40 patients, whose blood pressure was measured both manually, and by the computer. When the two sets of readings were compared, the level of accuracy was almost identical.
Dingchang is now seeking funds to turn his innovative solution into a prototype measuring device. This could form the basis of a future home testing kit, which anyone could measure their own blood pressure by connecting a traditional arm cuff to a computer, laptop, or smart-phone.
“I am confident that we can improve this 100-year-old acoustic technique,” Dingchang adds. “Machine learning is transforming our research. We can now teach computers to listen like a human ear. It is producing some of the most accurate automatic blood pressure measurements I have seen in my career.”
The Medical Device and Technology Research Group focuses on the development of innovative medical technologies and devices with scientific and socio-economic impacts to address unmet healthcare needs. The group works across multidisciplinary areas with electronic engineers, medical physicists, computer scientists, clinical consultants, industrial partners, guideline makers, and allied professionals at different stages along the pathway of medical device development and commercialisation.
Our group has established close research and clinical partnership with Essex Partnership University NHS Foundation Trust, and industry partners to combine academic research with commercial exploitation.