Template-Based Respiratory Monitoring and Tidal Volume Estimation

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Critical Care Medicine

Background:

• Traditional contact-based monitoring in ICUs has limitations.
• Vision-based respiratory monitoring shows potential but faces challenges like motion artifacts and region-specific analysis.
• Accurate respiratory volume estimation requires robust torso modeling.

Purpose of the Study:

• To develop and validate a novel vision-based approach for respiratory monitoring in ICUs.
• To improve the accuracy of respiratory volume and rate estimation.
• To address clinical constraints such as body motion and occlusion artifacts.

Main Methods:

• Utilized Kinect camera for capturing patient's body surface.
• Aligned and registered captured data with a pre-defined human template for consistent modeling.
• Implemented an efficient algorithm for volume computation and respiratory parameter extraction.

Main Results:

• Achieved over 90% accuracy in tidal volume estimation.
• Reached over 95% accuracy in respiratory rate estimation.
• Demonstrated robustness without requiring patient-specific calibration data.

Conclusions:

• The proposed vision-based system offers an effective and robust solution for ICU respiratory monitoring.
• The torso modeling approach enhances accuracy and practicality for clinical use.
• This technology presents a promising non-contact alternative to traditional monitoring methods.