Detection and staging of chronic obstructive pulmonary disease using a computed tomography-based weakly supervised...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Pulmonology

Background:

• Chronic obstructive pulmonary disease (COPD) is a significant global health issue, frequently underdiagnosed, especially in developing nations.
• Accurate and timely diagnosis of COPD is crucial for effective management and patient outcomes.
• Current diagnostic methods may not always be accessible or efficient for widespread screening.

Purpose of the Study:

• To develop and validate deep learning (DL) models for automated detection of spirometry-defined COPD using computed tomography (CT) imaging.
• To create a DL model capable of staging COPD severity according to the GOLD classification system.
• To assess the efficacy of a CT-based DL approach as a potential case-finding tool for COPD.

Main Methods:

• A large, heterogeneous dataset of 1393 participants was compiled, including CT scans and pulmonary function test data.
• An attention-based multi-instance learning (MIL) model was trained for COPD detection on 837 participants' CT scans.
• External validation was conducted using 620 low-dose CT (LDCT) scans from the National Lung Screening Trial (NLST) cohort, and a 3D residual network was used for GOLD stage classification.

Main Results:

• The attention-based MIL model demonstrated strong performance in COPD detection, achieving an AUC of 0.934 in the internal test set and 0.866 in the NLST cohort.
• The DL model accurately graded 76.4% of confirmed COPD patients according to the GOLD scale.
• These results indicate high accuracy in both identifying COPD and determining its severity using CT imaging.

Conclusions:

• The developed chest CT-based deep learning approach effectively identifies spirometry-defined COPD and categorizes patients by GOLD stage.
• This automated CT-DL method shows potential as an efficient and accessible tool for COPD diagnosis and staging.
• The findings suggest a valuable role for AI in improving COPD screening and management globally.