Shortly after the release of the first Apple Watch in 2015, several disgruntled customers started posting to social media channels about a major flaw in their brand new gadget. The smartwatch failed to read a pulse, couldn't detect contact with skin, and repeatedly prompted for the passcode — all because the users had a wrist tattoo. When placing the device on skin without any ink, it worked perfectly.
Other wearables, which use the same kind of optical heart rate tracking technology, have had similar issues with tattoos and even dark skin. Recently, we found that the Polar M430 running watch couldn't provide readings of heart rate for dark skin, which the company chalked up to a firmware problem it has since repaired.
Essential reading: Trying out LogicInk's UV-sensing tattoo
The myriad complexities of our skin - from skin tone and tattoos, to the presence of arm hair and sweat - can pose a significant challenge when it comes to wearable design. Ultimately, gaining a better understanding of how and why these variables affect data collection will lead to more accurate devices. To that end, we asked a couple of experts in the field of wearable biosensing about what properties of skin can interfere with a good reading, and how companies could potentially get around these problems.
Kunal Mankodiya, assistant professor of Electrical, Computer, & Biomedical Engineering at the University of Rhode Island and director of the Weareable Biosensing Lab, designs wearable medical devices for remote management of conditions like Parkinson's, post-traumatic stress disorder, and heart disease. For his research projects, any malfunction due to skin-related factors could be life-threatening or lead to misdiagnosis.
"My lab is problem-centric, so we work closely with physicians, psychiatrists, and physiotherapists to understand the medical problem in detail," says Mankodiya. "Then, we look at the wearable technology that could fit with and solve that problem.
Mankodiya also worked on a watch to monitor Parkinson's patients
He uses the same optical technology as most commercial smartwatches and fitness trackers to record heart rate in patients. Photoplethysmography involves shining light into the skin, which causes the photons to scatter and absorb inside the tissue. Photodetectors then measure the amount of light that exits the skin some distance away.
Since green light is strongly absorbed by red blood, optical heart rate monitors tend to use green LEDs. The amount of absorption is greater with more blood present - in other words, absorption increases when the heart pumps as opposed to between pumps.
Sounds simple enough, right? However, both skin and tissue are complicated and can vary from person to person. Anything that prevents enough light from reaching the photodetectors will reduce the signal-to-noise ratio and produce an unreliable reading.
"Physiological conditions are a factor in accuracy. For instance, sweat is not good because it puts a water layer between the LED and detector, which creates noise," says Mankodiya. "Darker skin and tattoos absorb more light, and that would require different parameters for the device to collect accurate data."
Pulse oximetry, which measures a person's oxygen saturation with a finger clip, utilizes photoplethysmography as well. Several studies have reported a drop in accuracy for those with darker skin. And Mankodiya has run into similar challenges with another type of optical technology for non-invasive brain monitoring called near-infrared spectroscopy (NIRS), where light sources and detectors are placed on the head. When using NIRS, his lab has found that both skin color and hair color affect the signal quality. For instance, Mankodiya and his colleagues often can't get a useable signal from Asian people with thick black hair since it absorbs too much light.
Sensors of all kinds - biometric, temperature, humidity, and others - are part of Silicon Labs' vast product portfolio. The semiconductor design company headquartered in Austin, TX has worked on optical heart rate sensors and wearable applications for the last two years, selling over 5 million sensors for integration into devices.
Sid Sundar, senior project manager at Silicon Labs, described some of the skin-related challenges faced by the company's engineers when designing optical heart rate sensors. Along with dark skin and tattoos, he mentioned body fat as a factor to consider for light absorption.
An extra layer of fat on the arm could affect the measurement since the blood vessel would be a greater distance from the photodetector. Ultimately, because of the heterogeneity of users, companies are forced to choose parameters in their sensors that would work best for the general public.
"The response of the skin and tissue can vary with different wavelengths, so short of testing the devices on every single user, you have to choose a color that is optimized for the largest possible number of users," says Sundar.
One possible solution is to use multiple wavelengths of light for an optical sensor. If one wavelength is being absorbed by melanin, for example, another wavelength that works better for a darker skin tone could turn on in its place. Sweat and arm hair can also pose complications, but these aren't as much of a concern to him as fit.
The lack of good contact between a sensor and the skin can lead to interference from ambient light and motion artifact. This is something we've seen wearable makers try to tackle. Some devices, like the Fitbit Ionic, protrude at the back to keep a tighter lock against the skin and reduct the amount of ambient light seeping in.
"When we talk to customers to discuss sensor performance, issues can be caused by fit or something fundamentally wrong with the algorithm or sensor," says Sundar. "It tends to be the former."
The next generation of scientifically precise wearables will likely contain multiple types of sensors, aggregating and cross-checking their data. For instance, heart rate could be measured by both optical and electrocardiography (ECG) technologies in a single wearable, such as in Huami's Amazfit Health Band, the first wrist-worn ECG tracker on the market.
"Having multiple sensors in a single device adds a lot of value. An electrical-based sensor has no problem with dark skin but can also more susceptible to sweat and hair," Sundar says. "Combining readings from multiple sensors provides an advantage if you can't get a good signal from one type of sensor."
Combining sensors will likely be one answer to the problem. For wearables to go beyond simple activity tracking to becoming tools for 24/7 health and disease management, accuracy is a must and cannot waver based on attributes like skin tone or hair color. Wearable designers and engineers face a difficult challenge, but with some creativity and lots of testing, skin-related attributes should start posing less of a problem - or at least a significant one - in the not-too-distant future.
How we test