Intuitive and Ubiquitous Fever Monitoring Using Smartphones and Smartwatches

TL;DR

This work tackles accessible fever monitoring by estimating core-body temperature from thermistors embedded in smartphones and smartwatches during brief forehead contact. It treats the device as a heat-transfer probe, using features from thermistor time series and capacitive touch data and a linearized Newton's law model to predict core temperature, achieving an MAE of $0.743^{\circ}F$ ($0.4^{\circ}C$) and a limit of agreement of $\pm 2.374^{\circ}F$ ($\pm1.3^{\circ}C$) with $R^2 = 0.837$ in lab validation, plus a small clinical pilot. The approach emphasizes active sensing, cross-device thermistor readings (via root access), and controlled contact geometry to reduce noise. If scaled, it could improve population-level fever data, enable remote screening, and reduce healthcare waste, informing epidemiology and remote care.

Abstract

Inside all smart devices, such as smartphones or smartwatches, there are thermally sensitive resistors known as thermistors which are used to monitor the temperature of the device. These thermistors are sensitive to temperature changes near their location on-device. While they are designed to measure the temperature of the device components such as the battery, they can also sense changes in the temperature of the ambient environment or thermal entities in contact with the device. We have developed a model to estimate core body temperature from signals sensed by these thermistors during a user interaction in which the user places the capacitive touchscreen of a smart device against a thermal site on their body such as their forehead. During the interaction, the device logs the temperature sensed by the thermistors as well as the raw capacitance seen by the touch screen to capture features describing the rate of heat transfer from the body to the device and device-to-skin contact respectively. These temperature and contact features are then used to model the rate of heat transferred from the user's body to the device and thus core-body temperature of the user for ubiquitous and accessible fever monitoring using only a smart device. We validate this system in a lab environment on a simulated skin-like heat source with a temperature estimate mean absolute error of 0.743$^{\circ}$F (roughly 0.4$^{\circ}$C) and limit of agreement of $\pm2.374^{\circ}$F (roughly 1.3$^{\circ}$C) which is comparable to some off-the-shelf peripheral and tympanic thermometers. We found a Pearson's correlation $R^2$ of 0.837 between ground truth temperature and temperature estimated by our system. We also deploy this system in an ongoing clinical study on a population of 7 participants in a clinical environment to show the similarity between simulated and clinical trials.

Figures (9)

• Figure 1: A box diagram showing the user interact and the flow of features into the model to make fever estimates.
• Figure 2: Example posture of user interaction. Box A highlights the region of contact between the user's forehead and device screen closer to the top half of the device, and box B highlights the "camera-grip" style where the four corners of the device are pinched between the user's fingers.
• Figure 3: A side-by-side comparison of the contact area of the screen for a human trial and a simulated water bag trial. This demonstrates a consistency between region of the screen and percentage of the screen in contact with the heat source in both trials.
• Figure 4: A distribution of all 51 samples from lab validation along the 4 feature axes. Rate of heat transfer is plotted against sous-vide set temperature (top left) to show the distribution of set temperatures captured. Orange points indicate a sample over the $100.4^{\circ}F$ threshold of a fever while blue points represent samples below this threshold. Location of contact (top right), percent of screen in contact (bottom right), and initial temperature of device (bottom left) are plotted against rate of heat transfer for all 3 plots to show each of these features influences the rate of heat transfer.
• Figure 5: The (left) correlation and (right) Bland-Altman plot for temperature estimates made by a quadratic regression on 51 simulated samples from $95^{\circ}F$ to $102.5^{\circ}F$ controlled by the sous-vide precision water heater. The lines on the Bland-Altman plot (right) show the mean and 95th-percentile limit of agreement.