Cine MRI-derived radiomics features indicate hemodynamic changes in the pulmonary artery

Abstract

Although cine MRI-derived radiomics features in the cardiac blood pool have been used to represent cardiac function and motion, the clinical relevance of radiomics features in the great vessels is still unknown. The aim of the present study was to test the hypothesis that cine MRI-derived radiomics features of the pulmonary artery (PA) can represent hemodynamic abnormalities in pulmonary hypertension (PH). With the approval of the institutional review board (IRB), 50 PH patients (21 males, 36–89 years old, diagnosed with right heart catheterization [RHC]) and 23 healthy volunteers (14 males, 26–80 years old) were retrospectively enrolled in this study. All participants underwent cardiac 4D flow and cine MRI (25 retrospective phases) at the right ventricular (RV) outflow tract (RVOT). A total of 93 radiomics features were extracted from RVOT cine images through a fixed size region of interest (ROI) at the proximal part of the PA. The peak values of the 6 first order features were different between the PH patients and controls. 4D flow-derived mean velocity in PA was related to ‘Kurtosis’ (r = 0.452,), ‘Range’ (r = 0.426), ‘Autocorrelation’ (r = 0.407), ‘Joint Average’ (r = 0.459), ‘Sum Average’ (r = 0.459), ‘High Gray Level Emphasis’ (r = 0.41), ‘Large Dependence High Gray Level Emphasis’ (r = 0.44), ‘High Gray Level Run Emphasis’ (r = 0.422), ‘Gray Level Variance’ (r = 0.419), ‘High Gray Level Zone Emphasis’ (r = 0.451), and ‘Small Area High Gray Level Emphasis’ (r = 0.415). Mean RV pressure was related to ‘Inverse Variance’ (r = 0.43) and ‘Run Percentage’ (r = 0.403). All p values < 0.05. Cine MRI-derived PA radiomics features have the potential to serve as novel imaging biomarkers for representing hemodynamic changes in pulmonary circulation.