Biomechanical Phenotyping of Forced Expiration for Precision Pulmonary Rehabilitation: A Machine Learning Approach to...

Area of Science:

• Pulmonary Biomechanics
• Respiratory Physiology
• Computational Biology

Background:

• Standard spirometry lacks mechanical dynamics of forced expiration.
• Novel biomechanical parameters are needed to predict respiratory impairments.

Purpose of the Study:

• Derive novel biomechanical parameters from spirometry.
• Establish functional respiratory phenotypes.
• Predict clinical respiratory impairments.

Main Methods:

• Analyzed 16,596 spirometry records (NHANES 2007-2012).
• Derived parameters for kinetic power, mass constraint, and airway instability.
• Applied principal component analysis, K-means clustering, and a neural network.

Main Results:

• Identified three phenotypes: Load-Constrained (45.4%), Mechanically Efficient (23.5%), Dynamic Collapse (31.0%).
• Aging reduced kinetic power, with steeper decline in males (p < 0.001).
• Neural network achieved 93.2% accuracy; Dynamic Airway Collapse Ratio, BMI, and kinetic power outperformed age and sex.

Conclusions:

• Biomechanical factors are superior predictors of pulmonary dysfunction over demographics.
• Phenotype classification enables targeted precision cardiopulmonary rehabilitation.