Pulmonary function testing in lung cancer: Analysis and implications

Abstract

Author(s): Rupa Mazumder, Sandeep Kumar C, Vijay Upadhye

Objective: International norms and professional judgment are the main foundations for the use of pulmonary function testing. The description of a common pattern is currently described using predetermined cut-offs. Based on the ATS/ERS (American Thoracic Society/European Respiratory Society) interpretation approach, we sought to investigate the anticipated illness outcome. Then, we looked at whether different decision trees that integrated lung function with clinical characteristics may lead to a more precise diagnosis using an impartial machine learning framework.

Materials and methods: Data from 968 participants who were first admitted to a pulmonary clinic were included in our research. Complete pulmonary function and studies that were chosen at the doctor's discretion formed the basis of the final clinical diagnosis. Clinical diagnoses were divided into ten categories and approved by a panel of experts.

Results: The ATS/ERS algorithm correctly diagnosed 38%. Only Chronic Obstructive Pulmonary Disease (COPD) was accurately diagnosed (74%). After 10-fold cross-validation, the new data-based decision tree raised detection accuracy to 68% for the most common lung disorders, with COPD, asthma, interstitial lung disease, and neuromuscular condition having considerably better positive predictive values and sensitivity.

Conclusion: Our findings demonstrate that computer-based selection of lung function and clinical variables and associated decision-making criteria may enhance the present algorithms for lung function interpretation.

Share this article