Resolving phenotypic and prognostic differences in interstitial lung disease related to systemic sclerosis by computed tomography-based radiomics

Abstract

Radiomic features are quantitative data calculated from routine medical images and have shown great potential for disease phenotyping and risk stratification in cancer. Patients with systemic sclerosis (SSc), a multi-organ autoimmune disorder, have a similarly poor prognosis (10-year survival of 66%), due to interstitial lung disease (ILD) as the primary cause of death. Here, we present the analysis of 1,355 stable radiomic features extracted from computed tomography scans from 156 SSc-ILD patients, which describe distinct disease phenotypes and show prognostic power in two independent cohorts. We derive and externally validate a first quantitative radiomic risk score, qRISSc that accurately predicts progression-free survival in SSc-ILD and outperforms current clinical stratification measures. Correlation analysis with lung proteomics, histology and gene expression data in a cross-species approach demonstrates that qRISSc reverse translates into mice and captures the fibrotic remodeling process in experimental ILD.