Improving Prognostication in Pulmonary Hypertension Using AI-quantified Fibrosis and Radiologic Severity Scoring at Baseline CT

• DOI: 10.1148/radiol.231718
• Publication Date: 2024-02-06T14:54:43Z
• Authors: Krit Dwivedi; Michael Sharkey; Liam Delaney; Samer Alabed; Smitha Rajaram; Catherine Hill; Christopher Johns; Alexander Rothman; Michail Mamalakis; A. A. Roger Thompson; Jim Wild; Robin Condliffe; David G. Kiely; Andrew J. Swift
• Keywords: Concordance

Background There is clinical need to better quantify lung disease severity in pulmonary hypertension (PH), particularly idiopathic arterial (IPAH) and PH associated with (PH-LD). Purpose To fibrosis on CT angiograms using an artificial intelligence (AI) model assess whether this approach can be used combination radiologic scoring predict survival. Materials Methods This retrospective multicenter study included adult patients IPAH or PH-LD who underwent incidental imaging between February 2007 January 2019. Patients were divided into training test cohorts based the institution of imaging. The cohort examinations performed 37 external hospitals. Fibrosis was quantified established AI radiologically scored by radiologists. Multivariable Cox regression adjusted for age, sex, World Health Organization functional class, vascular resistance, diffusing capacity lungs carbon monoxide performed. performance predictive models without AI-quantified assessed concordance index (C index). Results 275 (median 68 years [IQR, 60–75 years]; 128 women) 246 65 51–72 142 patients, respectively. analysis showed that percentage increased risk patient mortality (hazard ratio, 1.01 [95% CI: 1.00, 1.02]; P = .04). finding validated index, 0.76). combining improved predicting compared a including alone 0.67 vs 0.61; < .001). Conclusion Percentage survival when alone. © RSNA, 2024 Supplemental material available article.