Expiratory Flow and Volume Estimation Through Thermal-CO[Formula: see text] Imaging.

Area of Science:

• Biomedical Engineering
• Pulmonology
• Medical Imaging

Background:

• Current pulmonary function tests often require physical contact and can be influenced by patient effort.
• There is a need for non-contact, effort-independent methods to assess respiratory parameters.
• Visual analytics of physiological behaviors offer a novel approach to quantitative health monitoring.

Purpose of the Study:

• To introduce a quantitative, non-contact method for estimating exhale flow and volume using Thermal-CO2 imaging.
• To develop a system for fine-grain respiratory analysis based on visual analytics of exhale behaviors.
• To enable effort-independent pulmonary evaluation for natural breathing analysis.

Main Methods:

• Utilized CO2-filtered infrared visualizations of exhale behaviors.
• Obtained breathing rate, volumetric flow (L/s), and per-exhale volume (L) estimations.
• Developed two Long-Short-Term-Memory (LSTM) behavioral estimation models using per-subject and cross-subject training datasets.

Main Results:

• Per-individual model achieved R²=0.912 for flow correlation and 75.65-94.44% accuracy for in-the-wild volume.
• Cross-patient model demonstrated R²=0.804 for flow correlation and 62.32-94.22% accuracy for in-the-wild volume.
• Validated visual flow analysis for quantitative respiratory metrics.

Conclusions:

• The Thermal-CO2 imaging method enables non-contact estimation of exhale flow and volume.
• This approach facilitates effort-independent analysis of natural breathing patterns.
• Broadens capabilities for pulmonological assessment and long-term respiratory monitoring.