Smart Thermometer Can Tell Flu From Cold

Before the outbreak of coronavirus, the seasonal flu was one of the most dangerous infectious diseases, but a lot of people have trouble telling the difference between a flu and a cold by their symptoms alone. This gave [M. Bindhammer] the idea to design a smart thermometer that can distinguish between flu and cold.

Automated medical diagnostics is certainly an important technology of the future. [M. Bindhammer]’s project, named F°LUEX, is the second version of his iF°EVE thermometer. After taking the body temperature it asks the patient a set of questions about his symptoms and then calculates the probability of whether it is more likely to be a flu or a cold. [M. Bindhammer] uses a method commonly used in medical diagnostics based on Bayesian statistics which assigns a probability score to both hypotheses. It takes into account how often a certain symptom occurs when you have a common cold or flu as well as the overall probability of catching one or the other.

The hardware of the project is based on a custom PCB that includes a medical-grade MLX90614 infrared thermometer with an accuracy of ±0.2˚C around the human body temperature. The sensor is being read out by a Teensy 3.2 and information is displayed on a small OLED screen. Everything is housed in a 3D printed enclosure that received a nice finishing by painting with primer and acrylic spray paint. Unfortunately, [M. Bindhammer] project also got delayed by the corona crisis as his order for the temperature sensor got canceled due to the current high demand. But that does make us wonder how useful this could be to discriminate between cold, flu, and COVID-19.

An IR thermometer is something useful to have around not only for medical applications and can also be built without a custom PCB and minimal parts.

20 thoughts on “Smart Thermometer Can Tell Flu From Cold

      1. Conscientious people probably shouldn’t have their projects subjected to articles like this that make medical claims about them, though.

        I’ll guess we’ll find out the weight of that claim by how loudly he complains. Or if.

    1. to be honest, I am not sure why we don’t have such diagnostic tools already.

      this can be so useful and potentially provide low cost medical advise for the masses.

      I mean you need to be careful with the questions.
      However, statistically and procedural more people are going to get the same result and more consistent level of care.

      I am not saying we should get rid of doctors. I am saying this can be a useful tool and may lower medical cost.

      1. Ah yes, the blinding truth that there are large portions of the world that do not have access to this magical “cloud” that software folks are so lovingly embracing.

        Something small, runs on commonly available batteries, and being able to get something out that you haven’t even thought about for months, that will simply turn on and work is fantastic!

        I think this is worthy of a little praise for the developer and seems like it would have been an ideal candidate for one of the Hackaday Prizes.

  1. Really nice work here, something like this set up to differentiate between a good number of common infections could be really useful in controlling covid-19. Much more effective than creepy tracking apps or barbaric lockdowns, and with the bonus of not intruding on civil liberties at all. Keep up the good work.

  2. How do you calibrate your temperature sensor? In such applications mechanics of the enclosure and a proper calibration are crucial to provide accurate measurement. Mostly due to the temperature gradient of the sensor thermopile.
    I have some small experience with the Melexis sensors and out of the box they are only good on paper. Also, you need a model to estimate temperature (with these kind of sensors you usually do not measure temperature directly unless you want to wait for a stable reading for a looong time). I guess you know all this…

    1. As usual. Take a commercially available clinical thermometer as a reference, a correction function can then be derived from the data series of both measured temperatures. Since the temperature range is very small, the correction function is very precise.

      1. Can the device take several tests to improve accuracy. Like say over 2-12 hours? Would successive temperature readings improve accuracy? Did you also consider taking normal readings to increase accuracy?

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.