Validation of a new, minimally-invasive, software smartphone device to predict sleep apnea and its severity: transversal study

Overview

• Obstructive sleep apnea (OSA) is a common condition that can lead to cardiovascular problems and excessive daytime sleepiness.
• It is often underdiagnosed because the gold standard for diagnosis, polysomnography (PSG), can be difficult to access.
• The study evaluates the performance of an app called Apneal, which uses a smartphone's microphone, accelerometer, and gyroscope to estimate a patient's Apnea-Hypopnea Index (AHI), a key measure of OSA severity.

Plain English Explanation

Obstructive sleep apnea (OSA) is a condition where a person's breathing repeatedly stops and starts during sleep. This can be a serious problem, as it can lead to issues with the heart and excessive tiredness during the day. Unfortunately, OSA is often not diagnosed because the best way to test for it, called polysomnography (PSG), can be hard for people to access.

To help address this, researchers tested an app called Apneal, which can use the sensors in a smartphone to try to detect signs of OSA. The app records the sounds of a person's breathing and the movements of their body while they sleep, and then uses this information to estimate the Apnea-Hypopnea Index (AHI) - a measure of how severe the person's OSA is.

The researchers wanted to see how well this app-based approach worked compared to the traditional PSG method. They had 46 people use the Apneal app while also undergoing a PSG test, and then compared the results. The study found that the manual scoring (where a person looked at the app's data) was quite accurate in identifying people with moderate to severe OSA. The automatic scoring (where the app used a machine learning model to analyze the data) also showed promising results, though more testing is still needed.

Overall, this research suggests that using a smartphone app could be a useful way to help diagnose OSA, especially for people who have difficulty accessing the traditional PSG test. By making it easier to screen for OSA, this technology could help more people get the treatment they need.

Technical Explanation

The researchers conducted a proof-of-concept study to evaluate the performance of the Apneal app, which uses a smartphone's microphone, accelerometer, and gyroscope to estimate a patient's Apnea-Hypopnea Index (AHI). The study involved 46 adult patients who underwent in-hospital polysomnography (PSG) testing, the gold standard for diagnosing OSA.

First, the researchers performed a manual scoring step, where they analyzed the signals recorded by the Apneal app. For patients with an AHI greater than 15 events per hour, the manual scoring had a sensitivity of 0.91 and a positive predictive value (PPV) of 0.89. For patients with an AHI greater than 30 events per hour, the sensitivity was 0.85 and the PPV was 0.94.

The researchers then evaluated an automatic scoring method based on a sequential deep learning model that was developed by Apneal. This model was used to automatically detect respiratory events from the recorded signals. The results showed promising performance, with an Area Under the Receiver Operating Characteristic Curve (AUC-ROC) of 0.85 and an Area Under the Precision-Recall Curve (AUC-PR) of 0.94 for identifying patients with an AHI greater than 15. For patients with an AHI greater than 30, the AUC-ROC was 0.95 and the AUC-PR was 0.93.

These findings suggest that the Apneal app, using both manual and automatic scoring approaches, can accurately estimate a patient's AHI compared to the gold standard PSG. This could make it easier for people to be screened and diagnosed with OSA, potentially leading to earlier treatment and improved health outcomes.

Critical Analysis

The study provides promising results for the use of a smartphone-based app to estimate OSA severity, but it also has some limitations that should be considered.

First, the study was a single-center proof-of-concept, and the sample size was relatively small (46 patients). A larger, multi-center validation study involving a more diverse population with varying OSA severity would be needed to confirm the generalizability of the findings.

Additionally, the automatic scoring method using the deep learning model was only evaluated in this initial version (0.1) of the Apneal app. Further development and testing of the model's performance would be necessary to ensure its reliability and robustness.

It's also worth noting that the study did not compare the smartphone-based approach to other alternative methods for OSA screening, such as those using contactless polysomnography, data augmentation techniques, or oxygen desaturation detection. A broader comparative evaluation could help better understand the strengths and limitations of the Apneal app in the context of other emerging technologies for OSA diagnosis.

Overall, this study provides a solid foundation for further research and development of smartphone-based solutions for OSA screening and diagnosis. With continued refinement and validation, these types of approaches could potentially improve access to care and lead to earlier intervention for individuals with sleep disorders or Alzheimer's disease.

Conclusion

This proof-of-concept study demonstrates that a smartphone-based app, Apneal, can accurately estimate a patient's Apnea-Hypopnea Index (AHI) compared to the gold standard polysomnography (PSG) test. Both the manual scoring and automatic scoring methods using a deep learning model showed promising results in identifying moderate to severe obstructive sleep apnea (OSA).

If these findings can be replicated and expanded upon in larger, multi-center studies, the Apneal app could potentially help increase access to OSA screening and diagnosis, leading to earlier treatment and improved health outcomes for those affected by this common but often underdiagnosed condition.