Department of Diagnostic Imaging, Division of Neuroradiology, Room AG 31, Sunnybrook Health Sciences Centre, 2075 Bayview Ave, Toronto, ON, Canada M4N 3M5.
Radiology. 2010 Apr;255(1):142-53. doi: 10.1148/radiol.09091021.
To systematically evaluate the diagnostic benefits and inter- and intraobserver reliability of an incremental computed tomographic (CT) protocol in the confirmation of clinically suspected stroke, with combined imaging and clinical data as the reference standard.
Institutional review board approval was obtained, and participants gave informed consent. A total of 191 patients (mean age, 67 years +/- 16 [standard deviation]; 105 men) with strokelike symptoms of no more than 3 hours duration were recruited. Blinded review was performed by four readers with limited stroke imaging experience. Diagnostic confidence was recorded on a five-point scale. Logistic regression analysis was used to calculate the difference between the real and observed diagnoses, adjusting for confidence. Predictive effects of observed diagnostic performance and confidence score were quantified with the entropy r(2) value. Sensitivity, specificity, and confidence intervals were calculated while accounting for multiple reader assessments. Receiver operating characteristic (ROC) analyses, including area under the ROC curve, were conducted for three modalities in combination with confidence score. Inter- and intraobserver agreement was established with the Cohen kappa statistic.
The final diagnosis was infarct in 64% of the patients, transient ischemic attack in 18%, and stroke mimic in 17%. Large-vessel occlusion occurred in 70% of the patients with an infarct. Sensitivity for stroke determination with noncontrast CT, CT angiography, and CT perfusion increased by 12.4% over that with noncontrast CT and CT angiography and by 18.2% over that with only noncontrast CT for a confidence level of 4 or higher. The incremental protocol was more likely to enable confirmation of clinical stroke diagnosis (odds ratio, 13.3) than was noncontrast CT and CT angiography (odds ratio, 6.4) or noncontrast CT alone (odds ratio, 3.3), The area under the ROC curve was 0.67 for the combination of noncontrast CT and confidence score, 0.72 for the combination of CT angiography and confidence score, and 0.81 for the combination of CT perfusion and confidence score. Inter- and intraobserver agreement increased with progressive sequence use.
An incremental stroke protocol that includes CT perfusion increases diagnostic performance for stroke diagnosis and inter- and intraobserver agreement.
系统评估在以联合影像学和临床数据为参考标准的情况下，对临床疑似中风患者进行增量 CT 方案在确认中的诊断优势和观察者内及观察者间的可靠性。
本研究获得了机构审查委员会的批准，所有参与者均签署了知情同意书。共招募了 191 名症状持续时间不超过 3 小时的中风样症状患者（平均年龄 67 岁 +/- 16 [标准差]；105 名男性）。四名具有有限中风成像经验的观察者进行了盲法评估。诊断信心以五分制记录。使用逻辑回归分析计算真实诊断与观察诊断之间的差异，并进行置信度调整。通过熵 r(2) 值量化观察诊断性能和置信度评分的预测效果。在考虑多位观察者评估的情况下计算了敏感性、特异性和置信区间。对三种模态结合置信度评分进行了接收者操作特征（ROC）分析，包括 ROC 曲线下面积。采用 Cohen kappa 统计量评估观察者间和观察者内的一致性。
最终诊断结果为 64%的患者为梗死，18%为短暂性脑缺血发作，17%为中风模拟。70%的梗死患者存在大血管闭塞。非对比 CT、CT 血管造影和 CT 灌注联合使用的敏感性比非对比 CT 和 CT 血管造影分别提高了 12.4%和 18.2%，置信度为 4 或更高。与非对比 CT 和 CT 血管造影（比值比，6.4）或仅非对比 CT（比值比，3.3）相比，增量方案更有可能确认临床中风诊断（比值比，13.3）。非对比 CT 和置信度评分组合的 ROC 曲线下面积为 0.67，CT 血管造影和置信度评分组合的 ROC 曲线下面积为 0.72，CT 灌注和置信度评分组合的 ROC 曲线下面积为 0.81。随着序列使用的逐步增加，观察者间和观察者内的一致性也随之提高。
包含 CT 灌注的增量中风方案可提高中风诊断的诊断性能和观察者内及观察者间的一致性。
