Your Wearable Is Lying to You

A practice called biohacking is surging in popularity, in large part due to recent advances in sensors and other wearable technologies. Biohacking revolves around the idea of using data-driven insights into our bodies' functions to optimize health, performance, and overall well-being. With the rise of increasingly sophisticated sensors and wearables, individuals now have unprecedented access to real-time data about various physiological parameters.

These devices, ranging from fitness trackers to smart clothing embedded with biometric sensors, enable continuous monitoring of metrics such as heart rate, sleep patterns, calorie expenditure, and even stress levels. This wealth of data allows biohackers to gain deep insights into their bodies' responses to different stimuli, helping them tailor their lifestyles, diets, exercise routines, and even cognitive practices for maximum efficiency and effectiveness.

But the insights one can gather are only as good as the accuracy of the measurements. As the old saying goes: garbage in, garbage out. Unfortunately, some of our smart devices produce a whole lot of garbage. Measurements of calorie expenditure, in particular, are notoriously bad. And those inaccurate calculations can have some very real impact on the people that rely on them, ranging from taking on inappropriate exercise routines to eating poorly.

The reason for these poor estimates is that most modern devices, like smartwatches and fitness trackers, simply do not have the sensing equipment that is necessary to accurately determine the number of calories a person has burned. This can result in estimates being off by as much as 40 percent, despite the perceived accuracy of the specific numbers being provided by the device.

To get truly accurate measurements, the current gold standard method involves the use of a calorimeter. These instruments measure heart rate, respiration, and the volume of exhaled carbon dioxide. A typical wearable can measure heart rate well enough, but measuring respiration, and especially exhaled gasses is impractical. However, these are critical factors, without which accurate estimations of energy expenditure are not possible.

Or rather, they were not possible in the past. Researchers at IIT Gandhinagar, Carnegie Mellon University, and Cornell University found that by utilizing a thermal camera, they were able infer one’s pattern of respiration. The key to this was realizing that inhalation leads to evaporation around the lips and nostrils, and that produces a measurable heat signature. Furthermore, they found that by combining this respiration information with heart rate and body temperature (also captured by the thermal camera), they could make a very accurate prediction of the number of calories an individual has burned.

The system, called JoulesEye, leverages a machine learning algorithm to learn the association between the captured physiological parameters and energy expenditure. In a trial that included 54 individuals that were asked to run or cycle for 15 minutes, it was shown that JoulesEye could repeatedly estimate energy consumption with an error rate of just 5.8 percent, as compared to a calorimeter. This process currently requires that the camera be pointed at the user’s face for about 40 seconds, which is a pretty big ask for a commercial device, so the team is working to reduce that amount of time.

Thermal cameras can reasonably be incorporated into today’s devices like smartwatches without much trouble. But cameras like the one used in the study cost roughly $45, so it would add significantly to the cost of a wearable. As such, this technology might be restricted to more purpose-built devices that are of interest to the hardcore biohacker. The researchers are working to adapt JoulesEye such that it can use lower-cost thermal cameras, however, so it may become practical for more use cases in the future.