Quantitative in vivo microscopy of interstitial lung disease using deep learning and artificial intelligence-based endobronchial optical coherence tomography

Endobronchial optical coherence tomography (EB-OCT) is a high-resolution, bronchoscope compatible imaging technique with the potential to serve as a microscopic complement to HRCT in interstitial lung disease (ILD) for early microscopic diagnosis and therapeutic response assessment. However, integration of EB-OCT into clinical settings remains challenging due to the time- and labor-intensive qualitative evaluation and/or manual segmentation of large datasets consisting of tens of thousands of images acquired from individual patients. Here, we demonstrate the feasibility of a robust, computationally-efficient EB-OCT image analysis framework using artificial intelligence (AI) with deep learning architecture for rapid, automated ILD feature segmentation and volumetric quantification.