Early Detection of Qualitative Fibrosis Progression Using Hybrid HU-Z-score CT Analysis: Overcoming Limitations of Traditional Quantitative Methods

ABSTRACT Background Progressive pulmonary fibrosis carries poor prognosis despite availability of antifibrotics. Current progression criteria rely on functional decline (FVC ≥10% decline over 6-12 months), which detects disease worsening after significant structural damage. Previous quantitative CT (qCT) methods using fixed Hounsfield unit (HU) thresholds or volume-only measurements have shown inconsistent sensitivity for early progression. We hypothesized that a hybrid approach combining HU thresholding with Z-score normalization would detect qualitative progression (tissue densification) before quantitative territorial expansion. Methods We developed a novel hybrid CT analysis method integrating: (1) HU threshold-based fibrosis detection (>-600 HU), (2) Z-score normalization for severity stratification (mild Z=1-2, moderate Z=2-3, severe Z≥3), and (3) five clinical progression criteria including qualitative worsening (ΔZ-score ≥0.5). The method was validated in two ILD patients with serial CT at short intervals (3.5 and 10 months). Automated lung segmentation, fibrosis quantification, and clinical decision support were implemented in Python (scikit-image, SimpleITK, NumPy). Results In the index case (progressive COPD-fibrosis overlap, 3.5-month interval), traditional volume-based analysis showed minimal change (+1 mL, +2%), below significance threshold. However, the hybrid method detected significant qualitative progression: Z-score increased from 2.35 to 2.87 (+0.52 SD, p