NXP Software algorithms boost the accuracy of optical heartbeat monitors
How NXP Software algorithms boost the accuracy of optical heartbeat monitors from near-zero to 95%
NXP Software is well known as the leader in developing algorithms which improve speech quality in mobile phones. So how can our echo cancelling experience help to revolutionize heartrate monitoring from optical sensors in smartwatch applications? “Quite a lot”, according to Frans De Buys, Engineering Manager at NXP Software. “Better algorithms could take the current accuracy rate from ‘near zero’ to 95%”
That’s quite a claim. But first let’s take a step back and see what the US Food & Drugs Administration (FDA) have to say about Health & Fitness apps.
FDA Mobile Medical Applications Q&A
The FDA website reported in mid 2014: ‘‘Mobile applications (apps) can help people manage their own health and wellness, promote healthy living, and gain access to useful information when and where they need it. These tools are being adopted almost as quickly as they can be developed. According to industry estimates, 500 million smartphone users worldwide will be using a health care application by 2015, and by 2018, 50 percent of the more than 3.4 billion smartphone and tablet users will have downloaded mobile health applications. These users include health care professionals, consumers, and patients.’
The FDA goes on to say: ‘The FDA encourages the development of mobile medical apps that improve health care and provide consumers and health care professionals with valuable health information. The FDA also has a public health responsibility to oversee the safety and effectiveness of medical devices – including mobile medical apps.’
That statement shows a general approval of health apps from government, but wisely sounds a cautionary note about the need for regulation and, by implication, the quality of apps.
Of course, family doctors will have their own view of health apps.
Health Apps that make you stay fitter and feel better
In an article for UK quality newspaper The Guardian, Dr Clare Garenda, a General Practitioner (family physician), wrote: “As a GP, I understand the power of this new technology and I welcome the control it gives my patients – making shared decision-making and self-care a reality (and bringing fun into it). But I also am wary of overestimating the power of apps – either to change behaviour or improve health. Measuring bodily functions isn’t everything. What can be measured doesn’t always matter – a small drop or rise in blood pressure is neither here nor there. Doctors do more than process data – and while this technology is powerful and amazing, we must never forget that medicine is an art, as well as a science.”
This touches on an interesting point that is of relevance to readers of this blog: which is that if a diagnosis is going to be made on the strength of data – either by a physician or a smart device – then the data itself needs to be accurate.
Current heartbeat monitors are “way off” in terms of accuracy
Writing for US publication c/net (link below), journalist Sharon Profis discovered that the current optical sensors aren’t delivering accurate results: “As I stood on a treadmill in his office with an EKG machine connected to my chest, Dr. Jon Zaroff said my heart rate was 146. But as I glanced down at my Basis Carbon Steel device, with my heart rate quickly dropping, it still said I was at 93. Even after repeat tests done in a doctor’s office, the finding wasn’t a fluke – the band was way off.”
Profis continues: “I put five leading smart devices with heart rate monitors to the test, measuring their accuracy with an EKG and the help of Dr. Zaroff, a cardiologist at Kaiser Permanente medical center in San Francisco. You can find my results below, but it seems the optical sensing technology used in many of today’s new, wrist-based mobile heart rate monitors is sometimes inaccurate. That’s in comparison to time-tested EKG machines (or the heart rate monitors that emulate them), which sense the electrical impulses that trigger your heartbeats.”
Which brings us neatly back to the beginning of this story. Frans De Buys reckons that NXP Software expertise could bring optical heartbeat monitor accuracy up to 95%.
“The problem with optical sensors”, De Buys explains, “is that they’re kind of OK when you are sitting down … but as soon as you start moving, there’s a whole lot of false data being generated. That’s caused by factors like movement of the arm (which is where wristbands go) and localized skin changes, movement of the sensor over the arm, and data artefacts from the device itself.”
“But,” De Buys continues, “our speech and sensing algorithms have a track record of doing two things which can improve accuracy. First, our audio expertise means we’re great at removing artefacts and substracting confusing data from the equation. Secondly, our sensing experience means we’re good at taking data from multiple sources and extrapolating something meaningful from it. In the case of an optical heart monitor, we can use data from the smartphone’s accelerometer to constantly predict relevant changes that need to be made to the heartbeat reading. So we can sense when the person is moving, or even use sound sensing to identify if the person is jogging, and extrapolate a more accurate heartbeat reading.
Better data … and better integration.
“Of course,” De Buys explains, “providing more accurate data is exciting for health applications, but another strength we have at NXP Software is actually integrating those algorithms into the software stack of mobile devices. We’ve already integrated high-level algorithms into over 3 billion phones and wearables from the world’s leading device makers … plus we have technical teams in China, South Korea, India, USA and Europe to give our customers local support. So apart from giving the benefits of greater accuracy, we can also cut build costs and reduce the time to market for our customers.”
Looking to the future, De Buys says that once NXP Software has perfected optical heartbeat monitoring, the company will move on to optimizing the performance of SpO2 oxygen saturation and VO2 Max oxygen uptake features. “The end result will be better data, more meaningful health apps and more accurate diagnoses.”
US FDA : Mobile Medical Applications
The Guardian : 10 health apps to help you stay fitter and feel better.
c/net : Do wristband heart trackers actually work? A checkup.