Association of Radiologic PPFE Change with Mortality in Lung Cancer Screening Cohorts

A multi-institutional team led by researchers from University College London and the SUMMIT Consortium has published a landmark study demonstrating that AI-quantified progression of a specific lung abnormality strongly predicts mortality. The research, analyzing over 16,500 participants from two major lung cancer screening trials (NLST and SUMMIT), used an automated algorithm to measure annual changes in pleuroparenchymal fibroelastosis (PPFE)—a type of upper lobe scarring—on routine low-dose CT scans.

The findings revealed that progressive PPFE was independently associated with significantly increased mortality risk across both cohorts, with hazard ratios of 1.25 in NLST and 3.14 in SUMMIT. In the SUMMIT cohort specifically, progressive PPFE was also linked to 2.79 times higher rates of respiratory hospital admissions and increased antibiotic/steroid use. The automated quantification method proved consistent across different screening populations.

This represents a major advance in computational radiology, transforming incidental CT findings into actionable prognostic biomarkers. The study suggests that AI-powered analysis of routine screening scans could automatically flag patients at elevated respiratory risk, enabling earlier interventions. The methodology combines computer vision (for segmentation) with statistical modeling to extract clinically meaningful signals from imaging data that radiologists might otherwise overlook.