Can We Trust the Heart Rate Data on our Smartwatches?
I recently wrote a story about the effectiveness of oximetry in wearable devices, which became quite popular. Later I came across several scientific articles on the accuracy of heart rate monitors in smartwatches, the conclusions from which were quite unexpected. Since many people liked the story about oximetry, I decided to get wind of how heart rate technology works and share my findings with you. By the way, did you know that the world’s first wearable wire-free heart rate monitor was launched in 1982 by Polar? So, this feature has long been found in many watches and bands, and therefore, even more interesting to know how effective and accurate it is today.
The classic heart rate monitor is a sensor on the finger, which works on the following principle. A light-emitting diode (LED) emits light, which is registered on the other side of the finger by a photodiode, which takes readings at a frequency of several hundred times per second. LED is needed because the light flux from the finger is very weak and poorly registered by the photodiode. The device’s placement on the finger is because the pulse wave is very well reflected from the fingertip, and therefore can be clearly identified.
In developing smartwatches, the idea arose to use them, in particular, to measure heart rate. The main problem is that the watch or band is placed on the wrist, so the classical scheme of placement of sensors on both sides of the hand cannot be applied here. Therefore, the diodes are placed on one side, and the photodiode registers the light reflected from the wrist. In this case, a green light is used for emission, as it gives the optimal ratio of the depth of penetration into the skin and reflection. If we talk about blue light, it practically does not penetrate the skin. In turn, red and infrared light inpour very deeply and reach vascular channels in which pulse waves travel differently, bringing the noise to the measurement. That is why smartwatches use infrared light only when measuring heart rate in the background or during sleep.
The inaccuracy of measurement by registration of reflected light and human activity necessitates additional sensors and intelligent algorithms to improve measurement results.
In particular, data from the accelerometer and gyroscope are also taken into account when measuring heart rate. The accelerometer measures the overload acting on the watch from three axes. In turn, the gyroscope detects rotational movements. Thanks to these sensors, the change in the position of the watch in space are determined. These data must be taken into account because when you move your hand, high overloads can act on the watch, and the blood in the vessels can be significantly disturbed, which affects the accuracy of measurements.
If you have used a smartwatch a few times, you know that heart rate monitoring can be done in two modes — passive and active. In passive mode, the accuracy of measurement is much higher but requires you to meet certain conditions — be stationary during the measurement and fix the hand in the appropriate position. It is an entirely different situation when you move or do active exercise. The heart signal level becomes weak, and therefore, it becomes challenging to count each cardio pulse. Therefore, such a signal is considered in the frequency domain. To do this, developers use a mathematical Fourier transform to obtain a signal spectrum and find the “thread” of the pulse on the spectrogram. However, it is not always easy to identify such a thread due to the imposition of many unwanted noises (particularly from movements). Various filters are used to separate such noises, which, among other things, take into account the signals from the accelerometer and gyroscope.
Because human activities are diverse, it is challenging to develop a universal algorithm that will find the “thread” of the pulse. At the same time, specific patterns of human movement are repetitive and, therefore, predictable. For example, walking, running, cycling, etc. For such activities, the developers add in wrist-worn devices ability to choose different models of human behavior for the best measurement results. But there are nuances too. For instance, in the case of running, it is crucial how the person fastens the watch on the wrist, what is the quality of the surface on which the person runs, physiology, and so on.
Given the above, this is a rather difficult task — to develop an algorithm that optimally combines the ability to predict and assess the real conditions in which monitoring is carried out. To solve it, developers have to use artificial intelligence to generalize the various human states and predict the pulse, depending on the available sensor indicators. Datasets are used to train such a neural network. Different groups of people who simulate typical behavioral scenarios are involved in collecting such data. The obtained data is compared with accurate data from a chest heart rate monitor, which records the heart’s electrical activity. Therefore, if you need to get the most accurate data during training — it is better to use a chest heart rate monitor. By the way, Apple also notes that if you can’t get stable data, it’s better to use the Apple Watch wireless module to connect to an external heart rate monitor, such as a chest sensor.
To understand what other factors affect the quality of measurements, I looked through Apple’s recommendations for the proper use of their watches. Thus, the accuracy of measurement is affected by:
- Correct placement of the watch. The back of the Apple Watch should fit snugly. The Apple Watch strap should be tightened during training. The sensors will only work if the Apple Watch is located on top of the wrist.
- The amount of blood flowing to the vessels. The blood supply is markedly different for each person and may also depend on the environment. For example, when training in cold weather, the blood supply to the wrist’s skin may be so small that the heart rate monitor will not be able to measure it.
- Permanent or temporary changes in the skin. For example, the ink, contours, and saturation of some tattoos can block the passage of light to the sensor, which reduces the likelihood of obtaining correct results.
- The nature of human movements. Rhythmic movements, such as running or cycling, provide more accurate results than tennis or boxing, where athletes move irregularly.
So, as we can see, many factors affect heart rate measurement quality while using smartwatches or bands, in particular:
- human physiology;
- the nature of human activities;
- specifics of the used technology (measurement by registration of the reflected, instead of through light);
- the ability to use additional data from other sensors;
- the presence of intelligent algorithms that help cut off unnecessary noise.
Scientific research also shows that wrist-worn devices are not so accurate as the telemetry-based chest-strap monitors. Besides, in high-intensity exercise, the accuracy of all wrist-worn devices falls. You can read more about this study in the article “Accuracy of commercially available heart rate monitors in athletes: a prospective study.”
Another study looked at the functioning of the Apple Watch to monitor abnormal heart rhythms. As it turned out, among 264 patients who reported that their Apple Watch captured an anxious heart rhythm, only 30 patients (11.4%) received a clinically confirmed cardiovascular diagnosis.
Even if the patient is healthy, false triggers can still cause problems: they can cause stress and anxiety and push patients to seek unnecessary medical care. Even people who have no symptoms may feel the need to talk to their doctor about their watches’ abnormal readings. All this, according to scientists, can lead to the overuse of the health care system.
And now, let’s move on to the conclusions. No matter how old the heart rate measurement technology is, the data obtained by smartwatches cannot be considered 100% accurate. The main problem here is the specifics of the used technology and the limitations due to the placement of watches. Therefore, it is necessary to evaluate the received data critically and don’t distress if the results of measurements go beyond the permissible. And for serious training, where it is crucial to monitor the heart rate, it is better to use a chest sensor with an electrode that shows values as close as possible to those taken by medical heart rate monitors.
You just read another post from In Fitness And In Health: a health and fitness community dedicated to sharing knowledge, lessons, and suggestions to living happier, healthier lives.
If you’d like to join our newsletter and receive more stories like this one, tap here.
