Purpose: To assess the diagnostic performance of a deep learning algorithm in detecting early-stage pulmonary fibrosis on high-resolution computed tomography (HRCT).
Methods: A retrospective study of 800 HRCT scans from three tertiary centers was conducted. The algorithm, trained on 500 annotated cases, was validated against radiologist consensus for 300 independent cases. Sensitivity, specificity, and area under the ROC curve (AUC) were calculated.
Results: The algorithm achieved a sensitivity of 92.4%, specificity of 88.1%, and an AUC of 0.93 in detecting fibrotic changes. Interobserver agreement between the AI model and radiologists showed a κ = 0.86.
Conclusion: The proposed deep learning model demonstrates robust diagnostic capability in identifying early pulmonary fibrosis and could augment radiologist efficiency in clinical practice.

Keywords: Deep learning, pulmonary fibrosis, HRCT, artificial intelligence, diagnostic imaging