Radial artery phenotyping in systemic sclerosis through ultra-high frequency ultrasound: a radiomic approach

Objective: Systemic sclerosis (SSc) is a disorder characterized by a massive vascular involvement. Imaging biomarkers of vascular involvement in SSc may have potential clinical implications for prediction of the pathogenesis of vascular complications. This study is aimed at identifying possible patterns of vascular wall disarray and remodeling in radial arteries of SSc patients, by means of ultrahigh frequency ultrasound (UHFUS). Design and method: 5 end-diastolic frames of the right radial arteries of 41 patients with SSc and 41 healthy controls were obtained by VevoMD (70 MHz probe, FUJIFILM, VisualSonics, Toronto, Canada). 74 radiomic features and 4 engineered parameters were extracted: inner and outer layer thickness, and presence of adjunctive acoustic interfaces (triple signal). A feature selection algorithm was applied to reduce the number of features. The selected features were used to train classification model, using Linear Support Vector Machine (SVM). Results: The SVM classification model showed good performance (sensitivity = 0.63, specificity = 0.88, accuracy = 0.75, AUC = 0.75) to discriminate SSc patients from controls using fifteen selected features. Inner layer (208±61 vs 179±47 μm, p = 0.04) and outer layer thickness (104±22 vs 120±36 μm, p = 0.03) were significantly higher in SSc than in controls, triple signal pattern more frequent in patients (p = 0.002). Conclusions: Wall ultrastructure of radial arteries of SSc patients is altered: inner and outer layer thickened, showing frequently a triple signal pattern. Radiomic approach allow to distinguish between radial images from SSc patients and controls with a 75% accuracy.