LATE-BREAKING ABSTRACT: Artificial intelligence to improve the diagnostic power of complete pulmonary function tests

Introduction: The use of complete pulmonary function tests (CPFT) is primarily based on expert opinion and international guidelines. Current interpretation strategies are using predefined cut-offs for the description of a typical pattern. We aimed to explore the diagnostic power of CPFT based on the ATS/ERS interpreting strategy. Subsequently, we investigated whether an unbiased machine learning framework integrating lung function with clinical variables may provide alternative decision trees which result in a more accurate diagnosis. Methods: Our study included data from 968 subjects admitted for the first time to a pulmonary practice. A final clinical diagnosis was based on the combination of spirometry, resistance, lung volumes and diffusion (CPFT) with other investigations that were decided at the physician9s discretion. Clinical diagnoses were separated into 10 predefined disease groups and validated by an expert panel. Results: The ATS/ERS algorithm resulted in a correct diagnostic label in 38% of the subjects. COPD was detected with an acceptable accuracy (74%) whereas all other diseases were poorly identified. The new data-based decision tree improved the general accuracy to 68% after 10-fold cross-validation when detecting most common lung diseases, with a significant higher PPV and sensitivity for COPD, asthma, ILD and neuromuscular disorder [83% and 78%, 66% and 82%, 52% and 59%, 100% and 54%, respectively]. Conclusions: Our data show that the algorithms for lung function interpretation can be improved by a computer-based choice of lung function and clinical variables and their decision making thresholds. They may provide a first step to automated interpretation of lung function.