A Vision-based System for Breathing Disorder Identification: A Deep Learning Perspective.

Area of Science:

• Medical Technology
• Computer Vision
• Artificial Intelligence

Background:

• Remote patient monitoring is advancing with computer vision.
• Automated recognition of sleep breathing disorders is a complex challenge.
• Marker-free, non-intrusive approaches are under-explored for this condition.

Purpose of the Study:

• To develop a deep learning approach for classifying sleep breathing disorders.
• To utilize 3D camera depth maps for breathing analysis.
• To assess the viability of computer vision for remote sleep breathing assessment.

Main Methods:

• Deep learning architectures were employed for classification.
• Depth maps from 76 patients with breathing disorders were recorded using 3D cameras.
• Individual breathing events were classified as normal or abnormal.

Main Results:

• The system achieved 61.8% accuracy in classifying breathing events.
• Demonstrated the feasibility of using computer vision for breathing disorder detection.
• Highlighted the potential for remote and non-intrusive monitoring.

Conclusions:

• Computer vision and deep learning are viable tools for assessing breathing quality during sleep.
• The proposed approach offers a non-intrusive method for sleep disorder monitoring.
• Further research can enhance accuracy and expand applications in telemedicine.