Virtual patients for mechanical ventilation in the intensive care unit

Area of Science:

• Critical Care Medicine
• Biomedical Engineering
• Computational Physiology

Background:

• Mechanical ventilation (MV) is crucial in intensive care units (ICUs), but patient variability complicates management.
• Personalized MV settings are needed to improve patient outcomes and reduce costs.
• Developing accurate predictive models for lung mechanics is essential for optimizing MV therapy.

Purpose of the Study:

• To develop a generalized digital clone, or in-silico virtual patient, for predicting lung mechanics.
• To enable accurate, patient-specific predictions of responses to changes in MV settings.
• To facilitate optimized and personalized mechanical ventilation strategies.

Main Methods:

• A nonlinear hysteresis loop model (HLM) was developed to capture patient-specific lung dynamics.
• The virtual patient model was automatically generated using hysteresis loop analysis (HLA) from clinical ventilator data.
• Model performance was evaluated on data from 18 volume-control (VC) and 14 pressure-control (PC) ventilated patients.

Main Results:

• Virtual patient models accurately predicted lung mechanics for PEEP changes up to 12 cmH2O in both VC and PC cohorts.
• High R² values were achieved for predicting peak inspiratory pressure (PIP) and lung volume (VFRC) in VC patients (R²=0.86, R²=0.90).
• Similarly high R² values were observed for predicting peak inspiratory volume (PIV) and VFRC in PC patients (R²=0.86, R²=0.83), with small absolute errors.

Conclusions:

• The virtual patient model accurately captures and predicts nonlinear, patient-specific lung mechanics.
• The model's accuracy and versatility support its use in guiding personalized MV therapy.
• Mechanically and physiologically relevant virtual patients can optimize critical care management.