Ability of Radiomics Versus Humans in Predicting First?Pass Effect After Endovascular Treatment in the ESCAPE?NA1 Trial

Background First?pass effect (FPE), that is, achieving reperfusion with a single thrombectomy device pass, is associated better clinical outcomes in patients acute stroke. FPE therefore increasingly used as marker of and procedural efficacy. We aimed to evaluate the ability thrombus?based radiomics models predict undergoing endovascular compare performance experts nonradiomics thrombus characteristics. Methods Patients thin?slice noncontrast computed tomography angiography from ESCAPE?NA1 (Efficacy Safety Nerinetide for Treatment Acute Ischemic Stroke) trial were included. Thrombi manually segmented on all images. Data randomly split into derivation set included training validation subset an independent test set. Radiomics features extracted The machine learning compared 3 expert stroke physicians predicting using area under receiver operating characteristic curves. Results Thin?slice images 554 divided (training [n=388] [n=55]) (n=111). A model combination tomography, angiography, tomography–computed difference achieved highest (area curve, 0.74 [95% CI, 0.64–0.84]) prediction FPE. This was higher than mean curve (0.62 0.53–0.71], P =0.01 curves). also performed nonradiomics?based such volume permeability measurements ( <0.05). Addition type did not improve chosen Conclusion radiomics?based characteristics performs traditional imaging treated thrombectomy.