A Damaged-Informed Lung Ventilator Model for Ventilator Waveforms.

Area of Science:

• Physiological modeling
• Pulmonary mechanics
• Biomedical engineering

Background:

• Existing lung models struggle to capture complex pathophysiology and patient-ventilator interactions like ventilator-induced lung injury (VILI).
• Simple models lack the complexity to represent real-world pressure/volume signals, while complex models are often not estimable with clinical data.

Purpose of the Study:

• To develop a novel damaged-informed lung ventilator (DILV) model.
• To mathematize ventilator pressure and volume waveforms, incorporating lung physiology, mechanical ventilation, and their interactions.
• To create a model that can be estimated with clinical data for improved utility.

Main Methods:

• Developed the DILV model by starting with nominal waveforms and adding clinically relevant, hypothesis-driven features.
• Features added correspond to pulmonary pathophysiology, patient-ventilator interaction, and ventilator settings.
• Evaluated the model by estimating 399 breaths from controlled mouse data and uncontrolled human ICU data, with and without ventilator dyssynchrony.

Main Results:

• The DILV model parameters uniquely and reliably recapitulate waveform features.
• The model demonstrated flexibility in reproducing clinical and laboratory waveform variability.
• The DILV model achieved ≈12 times lower cumulative mean squares error compared to the single compartment lung model.
• Estimated parameters correlated with known measures of lung physiology, such as lung compliance.

Conclusions:

• The DILV model offers a significant improvement over traditional models for analyzing lung mechanics and ventilator interactions.
• The model's ability to provide high-fidelity estimates of lung state and VILI sources supports its use in clinical monitoring and research.
• This approach has the potential to enhance the management of VILI and acute respiratory distress syndrome.