Current and future contributions of AI to pulmonary function test interpretation, diagnostic approaches, and predictions of disease progression.

INTRODUCTION

Artificial intelligence (AI) methods - including machine learning, deep learning, and explainable AI - are increasingly applied to pulmonary function testing (PFT) to enhance interpretation, standardize procedures, and support clinical decision-making.

AREAS COVERED: This narrative review synthesizes evidence on AI contributions to diagnostic classification, quality control, signal analysis, and predictive modeling in PFT. AI systems have demonstrated improved accuracy and reproducibility versus unaided clinicians for tasks such as ventilatory pattern classification, small-airway dysfunction detection, methacholine challenge prediction, and automated spirometry quality assessment. Explainable AI methods increase clinician trust and interreader concordance. Key limitations include variable algorithm performance, limited interpretability, dependence on the quality and representativeness of training data, and insufficient integration of comprehensive clinical context. Priority development areas are assembling large, diverse, and standardized datasets aligned with American Thoracic Society (ATS)/European Respiratory Society (ERS) guidelines; implementing bias detection and mitigation strategies; defining clinically meaningful endpoints for validation; advancing explainability frameworks; and ensuring interoperability across PFT platforms.

EXPERT OPINION: Multidisciplinary collaboration among clinicians, engineers, and information technology specialists is essential for safe, ethical, and effective deployment. AI is a powerful adjunct for PFT interpretation but should remain a decision support tool pending rigorous real-world validation.