Automated airway quantification associates with mortality in idiopathic pulmonary fibrosis

Wing Keung Cheung, Ashkan Pakzad, Nesrin Moğulkoç, Sarah Needleman, Bojidar Rangelov, Eyjólfur Gudmundsson, An Zhao, Mariam Abbas, Davina McLaverty, Dimitrios Asimakopoulos, Robert Chapman, Recep Savaş, Sam M. Janes, Yipeng Hu, Daniel C. Alexander, John R. Hurst, Joseph JacobmedRxiv 2023 · DOI

Introduction

Idiopathic Pulmonary Fibrosis (IPF) is a chronic, progressive, and often fatal lung disease characterized by the scarring and stiffening of lung tissue. Currently, clinicians rely heavily on visual assessments of high-resolution computed tomography (CT) scans to identify "traction bronchiectasis"—the physical distortion and widening of airways caused by surrounding fibrotic tissue. While useful, visual scoring is inherently subjective and typically focuses on peripheral lung regions where damage is most apparent.

There is a critical need for objective, granular tools to quantify how IPF affects the entire airway tree. This study investigates whether automated, quantitative measurements of the proximal airways (generations 2 through 6) can provide more precise prognostic information than traditional clinical benchmarks.

Methods

The study involved an observational cohort of 90 IPF patients from Ege University Hospital, Izmir, Turkey, diagnosed between 2008 and 2015. Researchers utilized a two-stage computational pipeline to analyze CT data:

  1. Segmentation: Airway trees were mapped using a hybrid approach. This combined the established Pulmonary Toolkit (PTK) with a custom-built 2D dilated U-Net deep learning architecture, trained on 25 manually segmented airway trees from both healthy and IPF subjects.
  2. Quantification (AirQuant): The AirQuant tool calculated two primary geometric metrics across the 2nd to 6th generations of the airway:
  3. Segmental Intertapering: This measures the difference in median diameter between adjacent airway segments. In healthy lungs, airways narrow predictably; in fibrotic lungs, this tapering is often lost.
  4. Segmental Tortuosity: This evaluates the ratio of the measured segmental length against the direct end-to-end length, essentially measuring how "twisted" or "wiggly" the airway has become due to surrounding fibrosis.

Statistical analysis utilized Cox proportional hazards models to estimate mortality risk, adjusting for age, gender, smoking status, antifibrotic use, CT usual interstitial pneumonia (UIP) patterns, and lung function metrics like Forced Vital Capacity (FVC) and Diffusion Capacity of Carbon Monoxide (DLco).

Results

The analysis revealed that automated airway metrics are powerful independent predictors of survival.

Mortality Prediction

In multivariable Cox regression analyses (adjusted for FVC), both metrics showed significant associations with mortality:

  • Reduced Intertapering: Hazard Ratio (HR) = 0.75 (95% CI 0.66–0.85, p < 0.001).
  • Increased Tortuosity: Hazard Ratio (HR) = 1.74 (95% CI 1.22–2.47, p = 0.002).

These results remained significant when adjusted for DLco, indicating that these geometric changes provide prognostic value beyond established lung function tests.

Key Insights

  • Proximal Airway Significance: While IPF is often viewed as a peripheral disease, this study demonstrates that the larger, proximal airways (generations 2-6) are significantly affected and hold vital prognostic clues.
  • Geometry as a Biomarker: The physical distortion of the airway—its "wiggliness" and lack of tapering—serves as a surrogate for the total fibrotic burden.
  • Superiority to Visual Patterns: AirQuant metrics showed stronger associations with mortality than the presence of a visual UIP pattern, suggesting that automated tools can detect subtle structural changes that human observers might overlook.

Discussion and Conclusion

The study concludes that AirQuant-generated measures of segmental intertapering and tortuosity are independently associated with mortality in IPF patients. These findings suggest that abnormalities in proximal airway generations, which are not typically the focus of IPF assessment, have significant prognostic value.

Limitations

As a pilot study with a cohort of 90 patients, the findings require validation in larger, multi-center populations. Additionally, automated segmentation still faces challenges in cases of extremely severe disease or significant breathing artifacts, which can lead to data exclusion.

Future Directions

Automated airway analysis could refine patient selection for therapeutic trials by identifying those at higher risk of progression. Future research may also combine these airway metrics with other AI-driven assessments, such as vessel-related biomarkers, to create a more comprehensive profile of lung health.

Abbreviations

Table of abbreviations used in the study. Source, p. 4

Table of abbreviations used in the study.

Summary updated

Abstract

Abstract Objectives The study was to examine whether the airway metrics associate with mortality in IPF patients. Methods We performed an observational cohort study (n=90) of IPF patients identified from Ege University Hospital. An airway analysis tool AirQuant calculated median airway segmental intertapering and segmental tortuosity across 2 nd to 6 th generations of IPF airways. Intertapering measures the difference in median diameter between adjacent airway segments. Tortuosity evaluates the ratio of measured segmental length against direct end- to-end segmental length. Univariable linear regression analyses examined relationships between AirQuant variables, clinical variables and lung function tests. Univariable and Multivariable Cox proportional hazards models estimated mortality risk with the latter adjusted for patient age, gender, smoking status, antifibrotic use, CT usual interstitial pneumonia (UIP) pattern and either forced vital capacity (FVC) or diffusion capacity of carbon monoxide (DLco) if obtained within 3 months of the CT. Results No significant collinearity existed between AirQuant variables and clinical or functional variables. On univariable Cox regression analyses, male gender, smoking history, no antifibrotic use, reduced DLco, reduced segmental intertapering and increased segmental tortuosity associated with increased risk of death. On multivariable Cox regression analyses (adjusted using FVC), segmental intertapering (Hazard Ratio (HR)=0.75, 95% CI=0.66-0.85, p&lt;0.001) and segmental tortuosity (HR=1.74, 95% CI=1.22-2.47, p=0.002) independently associated with mortality. Results were maintained with adjustment using DLco. Conclusions AirQuant generated measures of segmental intertapering and tortuosity independently associate with mortality in IPF patients. Abnormalities in proximal airway generations, which are not typically considered to be abnormal in IPF, have prognostic value. Key Points AirQuant generates measures of segmental intertapering and tortuosity. Automated airway quantification associates with mortality in IPF independent of established measures of disease severity. Automated airway analysis could be used to refine patient selection for therapeutic trials in IPF.
BibTeX
@misc{cheung2023automated,
  author = {Wing Keung Cheung and Ashkan Pakzad and Nesrin Moğulkoç and Sarah Needleman and Bojidar Rangelov and Eyjólfur Gudmundsson and An Zhao and Mariam Abbas and Davina McLaverty and Dimitrios Asimakopoulos and Robert Chapman and Recep Savaş and Sam M. Janes and Yipeng Hu and Daniel C. Alexander and John R. Hurst and Joseph Jacob},
  title = {{Automated airway quantification associates with mortality in idiopathic pulmonary fibrosis}},
  year = {2023},
  doi = {10.1101/2023.03.29.23287913},
  howpublished = {medRxiv},
}
Figure