Artificial Intelligence-Powered Remote Monitoring of Patients with Chronic Obstructive Pulmonary Disease

Chronic obstructive pulmonary disease (COPD) is a major health concern, being the third leading cause of death globally after heart disease and cancer. Early detection of deterioration and timely treatment can prevent acute exacerbation of COPD (AECOPD), reducing its severity and the need for hospitalization. Remote patient monitoring (RPM) plays a crucial role in both inpatient and outpatient care for COPD. When combined with artificial intelligence (AI) technologies, RPM can anticipate exacerbations and enable early therapeutic intervention.

AI-Enabled Technologies for Remote Monitoring of COPD Patients

• Physical Activity Monitoring:
  • Accelerometers: Widely used in consumer wearables (e.g., wristbands, smartwatches) and smartphones, accelerometers are small-scale micro-electro-mechanical system devices. They can track body motions and detect activities of daily living (ADL) like sitting, walking, and sleeping. For example, a study presented an early RPM system with a small-waist-worn unit that processes and classifies physical activity data. Another study combined smartphone sensors (accelerometer, gyroscope, gravity sensor) using machine-learning algorithms to infer physical activities. These experiments showed an average accuracy of above 86% for classifying activities such as walking, running, sitting, etc.
  • Camera-Based Sensors: Camera-based sensors with action recognition techniques (e.g., convolutional neural networks) offer an unobtrusive way to remotely track ADL. With proper training and model tuning, they can achieve an accuracy of approximately 80%. Continuous ADL detection also helps monitor other diseases affecting ADL, like Alzheimer’s and Parkinson’s.
• Audio Symptoms Monitoring:
  • Cough Detection: Studies have developed systems for cough detection. One presented a real-time low-power wireless respiratory monitoring system with cough detection using a microphone and AI-based audio analysis. Another used artificial neural networks to detect cough after signal processing.
  • Wheeze Sound Analysis: Wheeze signals have sufficient information for patient categorization based on COPD severity. Machine learning algorithms like support vector machine have been used to build classifiers, achieving an average sensitivity of above 90% and an average accuracy of above 85%. Audio-based cough detection systems powered by AI are increasingly applied in RPM and important for studying other COPD-related audio symptoms.
• Environmental Sensors:
  • Air Quality Sensors: Many air quality sensors are available. Most are based on an infrared light source (laser) with a photodetector. They measure dust larger than 1 mm and provide particulate matter density readings.
  • DHT22 Sensor: A low-cost sensor for temperature and humidity monitoring. It uses a capacitive humidity sensor and a thermistor, achieving 0% – 100% humidity readings with 2% – 5% precision and ±0.5°C resolution for the temperature range from –40 to 80°C.
• Pulse Oximetry: A wearable finger pulse oximeter is used. It is a thin clip-like device placed on body parts like the finger, ear lobe, or foot (for infants). The principle is based on the change in light absorption during an arterial pulse. Two light sources (visible red-light spectrum – 660 nm and infrared spectrum – 940 nm) alternately illuminate the area. The microprocessor calculates the ratio of absorbed spectra and compares with a saturation value table to obtain pulse oximetry.
• Respiratory Rate:
  • ECG Sensor: Many studies show that respiratory rate and wave morphology can be approximated by ECG-derived respiration. This method detects frequency by measuring the size of R-wave in QRS signals, with a preliminary study showing over 97% precision.
  • Contactless Sensing (Doppler Radar): Transmits radio waves and senses chest-reflected signals. The chest wall movement is remotely monitored as a waveform. A time-domain autocorrelation model processes radar signals for rapid and stable respiratory rate estimation.

Predicting COPD Exacerbation with RPM

Most COPD exacerbation symptoms can be detected remotely using the described sensing technologies. For example, respiratory sensors capture vital signs, and pulse oximetry devices monitor physical activity. A previous study used respiratory signals and a machine-learning technique (decision tree forest) to predict early AECOPD, achieving detection accuracies of 78.0% (detected episodes) and 75.8% (reported exacerbations).

Prospect

AI technology has advanced rapidly and has promising applications in COPD. AI-assisted computed tomography (CT) and other imaging diagnoses help in accurate COPD diagnosis. CT scans measure airway diameter and tube wall thickness to identify the stage and clarify progression. Wearable and contactless devices will be further improved for more effective RPM. AI is expected to impact the COPD remote monitoring market. In the future, COPD patients will enjoy hospital-grade care at home or in outpatient facilities. Smart hospitals with intelligent monitoring devices will provide better care for patients with complex needs. Caregivers, assisted by AI tools, can plan and coordinate care efforts according to patient demands, leading to better outcomes at lower costs.

Conclusion

RPM is crucial for COPD patients. With sensor and AI technology development, it can reduce economic and medical burdens. AI-integrated remote monitoring systems analyze data from various devices to understand conditions, determine trends, and generate early warnings for effective patient monitoring and treatment.

This article is based on the research by Xuying Li, Hao-Peng Zhou, Zhi-Jun Zhou, Nan Du, Er-Heng Zhong, Ke Zhai, Nathan Liu, and Linfu Zhou. The study was published in the Chinese Medical Journal in 2021 (Volume 134, Issue 13). The DOI is 10.1097/CM9.0000000000001529.

Was this helpful?

YesNo

0 / 0