Automated 3D ιnterstitial lung disease εxtent quantification: performance evaluation and correlation to PFTs

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Healthcare
• Pulmonary Medicine

Background:

• Interstitial lung disease (ILD) diagnosis and management require accurate assessment of disease extent.
• Current semi-quantitative assessments by radiologists can be subjective and time-consuming.
• Computer-aided diagnosis (CAD) schemes offer potential for objective and efficient analysis of medical imaging.

Purpose of the Study:

• To evaluate the performance of a novel computer-aided diagnosis (CAD) scheme for detecting and quantifying reticular and ground glass patterns in ILD patients.
• To assess the reliability and accuracy of the CAD scheme compared to expert radiologists' assessments.
• To investigate the correlation between CAD-derived ILD pattern extent and pulmonary function tests (PFTs).

Main Methods:

• Evaluation of a CAD scheme on 37 volumetric chest CT scans from ILD patients.
• Reliability analysis using intraclass correlation coefficient (ICC) comparing CAD output to radiologist ratings.
• Accuracy assessment using receiver operating characteristic (ROC) analysis (Area Under Curve, Az) and correlation analysis with PFTs.

Main Results:

• The CAD scheme demonstrated substantial reliability for overall ILD extent (ICC=0.809) and distinct reticular patterns (ICC=0.806), and moderate reliability for distinct ground glass patterns (ICC=0.543).
• High accuracy was achieved in detecting total ILD (Az=0.950), reticular patterns (Az=0.920), and ground glass patterns (Az=0.883).
• Significant negative correlations were found between reticular pattern extent and key PFTs, mirroring radiologist findings.

Conclusions:

• The evaluated CAD scheme shows reliable and accurate performance in quantifying ILD patterns on chest CT.
• CAD-based quantification aligns with expert radiologists' assessments and correlates with PFTs.
• The CAD scheme shows potential as an imaging biomarker for ILD staging and patient management.