Pulmonary response prediction through personalized basis functions in a virtual patient model

Area of Science:

• Physiology
• Mechanical Ventilation
• Computational Modeling

Background:

• Recruitment maneuvers with positive-end-expiratory-pressure (PEEP) improve lung volume but lack standardized, patient-specific settings.
• Optimal PEEP levels fluctuate with patient condition, necessitating personalized monitoring.
• Current methods for determining PEEP are not fully standardized.

Purpose of the Study:

• To develop and validate physiologically relevant basis function sets for predicting lung elastance evolution in virtual patient models.
• To incorporate novel elements for modeling and predicting distension elastance.
• To improve the accuracy and robustness of mechanical ventilation (MV) lung mechanics predictions.

Main Methods:

• Examined 3 basis function sets (exponential, parabolic, cumulative) for predicting lung mechanics.
• Validated predictions against recruitment maneuver data from 18 volume-controlled ventilation (VCV) and 14 pressure-controlled ventilation (PCV) patients.
• Assessed prediction accuracy up to 12 cmH2O of added PEEP across various baseline PEEP levels.

Main Results:

• The exponential basis function demonstrated superior performance across VCV and PCV modes compared to other sets.
• Median absolute peak inspiratory pressure (PIP) prediction error was 1.63 cmH2O for VCV patients.
• Median peak inspiratory volume (PIV) prediction error was 0.028 L for PCV patients, with R² = 0.90-0.95 for distension prediction.

Conclusions:

• An exponential basis function best captures lung recruitment mechanics across different mechanical ventilation modes.
• This model accurately predicts distension mechanics within 5-10% accuracy for the first time.
• The validated digital twin model enhances the prediction of lung injury risk before altering ventilator settings.