Receiver-Operator Curve (ROC) models for CoRad-rAD.

ROC analyses presented are for an upper limb (UL) score of ≤ 1 at day 7 and a modified Rankin Scale (mRS) score of ≤ 1 and ≤ 2 at three months. The optimal threshold of CoRad-rAD (compromise between specificity and sensitivity) for each model is given with the associated area under the curve (AUC), accuracy (Acc), specificity (Spec), and sensitivity (Sens). For each model, the upper bound of 100% negative predictive power (NPV) and the lower bound of 100% positive predictive value (PPV) are given.Receiver-Operator Curve (ROC) models for CoRad-rAD.

本次展示的受试者工作特征曲线（Receiver-Operator Curve，ROC）分析，针对两类评价场景：一是第7天上肢（upper limb，UL）评分≤1分，二是3个月时改良Rankin量表（modified Rankin Scale，mRS）评分≤1分与≤2分。针对每个模型，我们给出了CoRad-rAD的最优阈值（兼顾特异度与灵敏度的平衡点），以及对应的曲线下面积（area under the curve，AUC）、准确率（accuracy，Acc）、特异度（specificity，Spec）和灵敏度（sensitivity，Sens）。针对每个模型，同时提供了100%阴性预测值（negative predictive power，NPV）的上限，以及100%阳性预测值（positive predictive value，PPV）的下限。针对CoRad-rAD的受试者工作特征曲线模型。