Estimating a Physiological Lung Function Score and Biological Sex Using Pulmonary Function Tests and Machine...

Area of Science:

• Pulmonary Medicine
• Artificial Intelligence
• Biostatistics

Background:

• Sex and age are known determinants of lung function.
• Biological and anatomical factors influence sex- and age-related pulmonary differences.

Purpose of the Study:

• To develop a machine learning model for predicting lung age and sex using pulmonary function test data.
• To assess the model's performance and interpretability in estimating physiological lung age.

Main Methods:

• Retrospective analysis of pulmonary function tests from 6392 healthy adults.
• Training gradient-boosted machine models to predict chronological age and biological sex.
• Utilizing Shapley additive explanation values and partial dependence plots for model interpretability.
• Employing quantile regression to estimate reference percentiles for predicted lung age.

Main Results:

• The best age prediction model achieved a root mean square error of 7.01 years.
• The best sex classification model achieved an area under the curve of 0.981.
• Key predictors for lung age included residual volume, forced expiratory volume in 1 second, and alveolar volume.
• Peak expiratory flow, height, and age were influential features for sex classification.

Conclusions:

• Artificial intelligence can predict patient sex and estimate lung age from pulmonary function data.
• Validated AI algorithms for physiological lung age estimation may serve as a measure of respiratory health.