Fig. 3From: Automatic segmentation of the great arteries for computational hemodynamic assessmentSegmentation metrics and flow field errors. Three segmentations are compared in a pairwise fashion: machine learning (ML), ground truth (GT) and second observer (SO). A, B Confusion-based similarity metrics: dice score and IoU. C, D Distance-based similarity metrics: Hausdorff distance (HD) and average surface distance (ASD), measured in pixels. E, F CFD-derived pressure and velocity mean average percentage errors (MAPE)Back to article page