Integration of AI and Iot Technologies for Non-Invasive Sleep Apnea
Detection and Monitoring
Patrice Adlino, Parth Jadhav, Pavan Pawar, Amey Prayag
Department of Mechanical Engineering, Vishwakarma Institute of Technology, Pune
Received: 04 December 2025; Accepted: 09 December 2025; Published: 20 December 2025
ABSTRACT
The prevalence of sleep apnea has driven demand for low-cost, non-invasive, and continuous home-based
monitoring systems. This research paper presents an AI-enabled IoT architecture integrating wearable sensors,
ESP32 microcontroller processing, and cloud-based analytics for real-time detection of apnea events. The system
monitors physiological parameters including SpO2, heart rate, respiratory rate, and body movement. Machine
learning models such as SVM, Random Forest, CNN, and LSTM enhance detection accuracy. Findings
demonstrate 93% accuracy and 94% sensitivity, validating the effectiveness of the system as a scalable
alternative to clinical polysomnography.
INTRODUCTION
Sleep apnea is a chronic disorder characterized by repeated interruptions in breathing during sleep, leading to
oxygen desaturation, cardiovascular stress, and long-term health risks. Conventional diagnosis relies on clinical
polysomnography (PSG), which is accurate but expensive, time-consuming, and inaccessible to many patients.
This study proposes a non-invasive, AI-powered IoT system designed to detect sleep apnea events autonomously
through real-time monitoring. The solution emphasizes low cost, portability, remote accessibility, and continuous
data analytics suitable for home-based monitoring.
LITERATURE REVIEW
Recent advancements in smart healthcare highlight the integration of IoT and AI for physiological monitoring.
Alshehri & Muhammad (2021) provide a broad survey of IoT-driven healthcare, emphasizing remote monitoring
benefits. Olsen et al. (2024) demonstrate feasibility in using wearable sensors for sleep stage detection via deep
learning. Systems such as SHIFT and Fusion-Infused Hypnocare exhibit the growing trend toward multimodal
real- time health monitoring.
Explainable AI models like Random Forest and deep-learning frameworks such as CNNs and LSTMs have
proven effective in identifying respiratory anomalies. These studies validate the potential for decentralized,
automated, and continuous detection systems for sleep apnea, aligning closely with the system developed in this
work.
METHODOLOGY
The methodology includes hardware integration, data acquisition, signal preprocessing, AI model development,
and cloud-based analysis. Wearable sensors—MAX30102 pulse oximeter, respiratory monitor, and MPU6050
accelerometer—collect key physiological signals.
The ESP32 microcontroller handles preliminary filtering and feature extraction. Cleaned data is transmitted via
Wi- Fi or Bluetooth to a cloud server for real-time processing. Machine learning models classify apnea events
based on thresholds and pattern detection.
Page 1765
