Quantitative CT Analysis in Chronic Hypersensitivity Pneumonitis: A Convolutional Neural Network Approach

Area of Science:

• Pulmonary Medicine
• Radiology
• Artificial Intelligence in Medicine

Background:

• Chronic hypersensitivity pneumonitis (cHP) is a complex interstitial lung disease (ILD) with variable presentations, including small airway involvement and fibrosis.
• Accurate characterization of lung tissue in cHP is crucial for understanding disease progression and patient outcomes.
• Computer-aided analysis, particularly using convolutional neural networks (CNNs), offers potential for objective quantification of ILD features on CT imaging.

Purpose of the Study:

• To quantify the extent of various pathological classes (consolidation, ground glass opacity, fibrosis, low attenuation areas, reticulation, healthy regions) in cHP patients using a CNN.
• To determine the correlation between these quantified pathological extents and pulmonary function tests (PFTs).
• To investigate the relationship between pathological patterns identified by CNN and the presence of mosaic attenuation on CT scans.

Main Methods:

• High-resolution computed tomography (HRCT) scans from 27 cHP patients were analyzed.
• A CNN method was employed to quantify the extent of six textural features: consolidation (C), ground glass opacity (GGO), fibrosis (F), low attenuation areas (LAA), reticulation (R), and healthy regions (H).
• The quantified extents were correlated with PFTs (FVC%, FEV1%, TLC%, DLCO%) and the presence of mosaic attenuation.

Main Results:

• Healthy lung regions (H) positively correlated with all measured PFTs (FVC%, FEV1%, TLC%, DLCO%).
• Ground glass opacity (GGO), reticulation (R), and consolidation (C) showed negative correlations with FVC% and FEV1%, with reticulation exhibiting the strongest negative correlation.
• Fibrosis (F) negatively correlated with DLCO%. Patients with mosaic attenuation patterns exhibited significantly more healthy regions (H) and less reticulation (R) and consolidation (C), along with better lung function indices (FVC%, DLCO%).

Conclusions:

• CNN-based quantification of pathological tissue extent in cHP provides improved characterization and correlates significantly with PFTs.
• The study suggests that mosaic attenuation patterns in cHP primarily represent patchy ILD rather than small airway disease.
• CNN analysis offers a valuable tool for objective assessment and characterization of cHP, potentially aiding in disease management and prognosis.