Application of Artificial Intelligence in the Interpretation of Pulmonary Function Tests

Area of Science:

• Pulmonology
• Medical Informatics
• Artificial Intelligence

Background:

• Chronic obstructive pulmonary disease (COPD) and asthma represent a significant global health burden, with COPD being a leading cause of death worldwide and in India.
• Pulmonary function tests (PFTs), primarily spirometry in resource-limited areas, are crucial for diagnosing respiratory conditions but suffer from subjective interpretation and inter-observer variability.
• Artificial intelligence (AI) offers a potential solution to enhance the accuracy and consistency of PFT interpretation.

Purpose of the Study:

• To compare the diagnostic accuracy of AI algorithms against senior pulmonologists in interpreting PFTs using limited clinical data and spirometry.
• To assess the consistency and accuracy of pulmonologists' interpretations when faced with the same patient data.

Main Methods:

• AI algorithms were developed and evaluated.
• Spirometry and limited clinical data from 440 patients were analyzed by both an AI algorithm and eight senior pulmonologists.
• Key performance metrics including accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated.

Main Results:

• The study included a majority of male patients (approx. 60%) aged 21-60 (approx. 70%).
• Pulmonologists achieved an average accuracy of 65.82% against a gold standard, with a Fleiss's kappa of 0.46 indicating moderate agreement.
• The AI algorithm demonstrated a significantly higher accuracy of 86.59%.

Conclusions:

• Pulmonologist interpretation of PFTs with limited data exhibits lower accuracy and considerable variability.
• AI algorithms provide consistent and high accuracy in PFT interpretation, surpassing human performance in this study.
• Implementing AI in clinical practice, particularly in resource-constrained regions, can significantly improve diagnostic quality and reduce inter-observer variability in respiratory medicine.