JMIR Publications

ABSTRACT

Background:

Spirometry can be performed in the office setting or even remotely from portable spirometers. Although basic spirometry can be diagnostic for obstructive lung disease, clinically pertinent information such as restriction, hyperinflation, and air-trapping require additional testing, such as body plethysmography, which is not as readily available. We hypothesize that spirometry data contains information that will allow estimation of static lung volumes in certain circumstances, leveraging machine learning techniques.

Objective:

To develop AI algorithms for estimating lung volumes from spirometry measures.

Methods:

This study obtained spirometry and lung volume measurements from the Mayo Clinic pulmonary function test database for patient visits ranging from February 19th, 2001, through December 16th, 2022. Pre-processing was performed and various machine learning algorithms, including a generalized linear model with regularization, random forests, extremely randomized trees, gradient-boosted trees, and XGBoost for both classification and regression cohorts.

Results:

A total of 121,498 PFTs were used in this study, with 85,017 allotted for exploratory data analysis and model development and 36,481 tests reserved for model evaluation. The mean age across the cohort was 64.7 years (range 18 – 119.6), with a balanced distribution between genders (48.2% female and 51.8% male). The classification cohort results showed a strong performance overall, with relatively low RMSE and MAE values observed across all predicted lung volumes. Across all lung volume categories, the models demonstrated strong discriminatory capacity, as indicated by high AUC values ranging from 0.85 to 0.99 in the training set and 0.81 to 0.98 in the testing set.

Conclusions:

Overall, the models demonstrate robust performance across lung volume measurements, underscoring their potential utility in clinical practice for accurate diagnosis and prognosis of respiratory conditions in locations where access to body plethysmography or other lung volume measurement modalities is challenging.