ENHANCING PULMONARY FUNCTION TEST REPORTING WITH ARTIFICIAL INTELLIGENCE: A RETROSPECTIVE OBSERVATIONAL STUDY

Abstract

Background: Pulmonary Function Tests (PFTs) are essential for diagnosing and monitoring respiratory disorders such as asthma, chronic obstructive pulmonary disease (COPD), and interstitial lung disease. Despite standardized protocols, manual interpretation of PFTs is prone to interobserver variability and delays in reporting. Artificial Intelligence (AI) offers a promising solution to improve diagnostic accuracy, consistency, and efficiency in functional respiratory diagnostics. This study aims to evaluate the accuracy and reliability of AI-assisted interpretation of PFTs compared with conventional assessment by pulmonologists.

Methods: A retrospective observational study was conducted at the Pulmonary Function Laboratory, Saveetha Medical College, Chennai, including 200 patients who underwent spirometry. Data were processed using a machine learning based algorithm trained on a subset of cases to classify ventilatory patterns into normal, obstructive, restrictive, and mixed categories. The diagnostic performance of AI interpretation was validated against pulmonologist reports. Key outcome measures included sensitivity, specificity, predictive values, and concordance with expert assessment.

Results: Among the study participants, the majority were male (86%), with most falling within the 20–39-year age group (51.5%). The AI model demonstrated strong diagnostic accuracy across all categories. For mixed obstructive patterns, the area under the curve (AUC) was 0.901; for obstructive, 0.930; for restrictive, 0.930; and for normal spirometry, 0.959. Sensitivity and specificity consistently exceeded 84% and 95%, respectively, with high positive and negative predictive values. The AI system reduced interobserver variability and produced consistent, reproducible outputs comparable to pulmonologist interpretations.

Conclusion: AI-based interpretation of PFTs achieves high diagnostic accuracy and reliability, offering a practical alternative to manual reporting. Its integration into clinical workflows and electronic health record systems can improve efficiency, reduce reporting delays, and enhance diagnostic consistency, particularly in resource-limited healthcare settings. Further validation across diverse populations is warranted before widespread adoption.