# Detection of sleep apnea using smartphone-embedded inertial measurement unit

## TL;DR

This paper shows that smartphones can detect sleep apnea using built-in sensors when worn on the abdomen during sleep.

## Contribution

The study demonstrates that non-wristwatch devices like smartphones can detect sleep apnea using embedded inertial sensors.

## Key findings

- IMU-derived respiratory events showed strong agreement with PSG apnea-hypopnea events, with F1 scores up to 0.821.
- Regression models using IMU signals correlated highly with PSG AHI (r up to 0.96).
- Moderate-to-severe sleep apnea was identified with high accuracy (AUC up to 0.98).

## Abstract

We previously demonstrated that sleep apnea (SA) can be detected using acceleration and gyroscope signals from smartwatches. This study investigated whether an inertial measurement unit (IMU) embedded in non-wristwatch devices, such as smartphones, can also detect SA when worn during sleep. During polysomnography (PSG), subjects wore an IMU-embedded GPS device (Amue Link®) and/or smartphones (Xperia® or iPhone®) on their abdomen. Triaxial acceleration and gyroscope signals were recorded overnight. Data were split into training and test groups (2:1) for each device. An algorithm was developed in the training groups to extract respiratory movements (0.13–0.70 Hz) and detect respiratory events, which were validated in the test groups. IMU-derived respiratory events showed breath-by-breath concordance with PSG apnea-hypopnea events, yielding F1 scores of 0.786, 0.821, and 0.796, respectively. Regression model derived from IMU signals correlated with PSG AHI in the test groups (r = 0.90, 0.93, and 0.96), with limits of agreement of -16.7 to 25.9, -17.4 to 22.5, and − 18.4 to 20.5. Using cutoff values from the training groups, moderate-to-severe SA (AHI ≥ 15) was identified in the test groups with AUCs of 0.95, 0.98, and 0.94 and F1 scores of 0.89, 0.96, and 0.92, respectively. IMUs embedded in non-wristwatch devices, including smartphones, can quantitatively detect SA when worn during sleep.

The online version contains supplementary material available at 10.1038/s41598-025-99801-3.