Contactless Respiratory Waveform Estimation Using a Depth Camera and AI-Based Body Detection

Area of Science:

• Medical Imaging and Sensing
• Artificial Intelligence in Healthcare
• Patient Monitoring Technologies

Background:

• Continuous patient observation during computed tomography (CT) is challenging, especially for adverse events like contrast media reactions.
• Existing monitoring methods can be intrusive or difficult to implement during CT procedures.

Purpose of the Study:

• To propose and evaluate a novel contactless method for respiratory waveform estimation during CT examinations.
• To assess the feasibility of using a depth camera and AI for non-invasive patient monitoring.

Main Methods:

• A contactless system employing a depth camera and AI-based body detection was developed for respiratory monitoring.
• The system extracts depth-based motion signals from anatomically relevant respiratory regions (ROIs).
• Performance was validated against a wearable respiration belt using quantitative (error metrics, correlation, Bland-Altman) and qualitative analyses.

Main Results:

• The proposed method achieved stable respiratory waveform estimation.
• The chest region demonstrated the lowest waveform error and highest correlation.
• Bland-Altman analysis revealed minimal systematic errors in respiratory rate, though variability was influenced by ROI and clothing.

Conclusions:

• Contactless respiratory monitoring during CT is feasible using depth cameras and AI.
• Careful selection of anatomical regions of interest is crucial for robust waveform extraction.
• This technology has the potential to enhance patient safety during CT scans.