
Contact the VUNO team anytime.
ECR 2026
A Retrospective Longitudinal Cohort Study
Background & Purpose Idiopathic Pulmonary Fibrosis (IPF) on Chest radiography (CXR) is one of the diseases that is difficult to diagnose early, and most patients are diagnosed only after the disease has progressed considerably. This study aimed to evaluate the potential of artificial intelligence (AI) in detecting early radiographic signs suggestive of IPF by retrospectively analyzing longitudinal CXR data. We specifically assessed the temporal gap between the AI-detected abnormality and conventional clinical decision points, including pulmonologist interpretation and final diagnosis, in a triage setting.
Methods Single-center retrospective cohort study analyzing 173 patients with CXRs flagged positive for "interstitial opacity" by VUNO Med®–Chest X-ray. Time-to-event analyses measured: (1) time from first CXR to AI First detection, and (2) time from AI First detection to confirmed diagnosis using Kaplan-Meier survival analysis, log-rank, and Cox proportional hazards models were used.
Key Findings AI detected abnormalities in most cases at initial CXR examination of serial CXRs. Median time from AI first detection to clinical diagnosis was 1,407 days (3.9 years). At 6 months post-AI first detection, 88.8-96.4% remained undiagnosed; at 1 year, 75.5-87.0% remained undiagnosed. Pre-2018 diagnoses showed significantly reduced diagnostic delay compared to post-2018. Patients with higher AI scores at the time of initial detection were diagnosed with IPF significantly earlier (median: 1,185 days) compared to those with lower scores (median: 1,628 days), although the difference did not reach statistical significance (log-rank p = 0.0705). This result suggests that AI scores may reflect the clinical severity or urgency perceived by pulmonologists, thereby facilitating earlier diagnostic work-up in high-score cases.