基于CT平扫的放射组学列线图鉴别动静脉畸形脑出血与原发性脑出血的价值

The value of radiomics nomogram based on CT in differentiating arteriovenous malformation cerebral hemorrhage from primary cerebral hemorrhage

目的开发基于基线平扫CT的放射组学列线图模型以鉴别动静脉畸形(AVM)脑出血与原发性脑出血。方法回顾性分析苏州大学附属第一医院2017年10月至2020年9月经手术证实的脑出血患者135例,其中AVM脑出血患者52例,原发性脑出血患者83例。从基线平扫CT提取放射组学特征,计算放射组学评分(Radscore)并构建放射组学模型;结合临床特征和CT征象的多元logistic回归分析用于建立临床模型,并结合Radscore建立列线图模型。使用ROC曲线和决策曲线分析(DCA)评估模型的判别性能。结果经筛选得到的6个特征用于建立放射组学模型。临床模型由年龄(OR 4.739,95%CI 1.382~16.250)和血肿位置(OR 0.111,95%CI 0.032~0.385)构成,列线图模型由年龄、血肿位置和Radscore构成。在训练组,列线图模型(曲线下面积为0.912)与临床模型(曲线下面积为0.816)、放射组学模型(曲线下面积为0.857)比较,差异有统计学意义(Z值分别为2.776、2.034,P值分别为0.006、0.042);在验证组,列线图模型(曲线下面积为0.919)与临床模型(曲线下面积为0.788)和放射组学模型(曲线下面积为0.810)比较差异无统计学意义(Z值分别为1.796、1.788,P值分别为0.073、0.074)。DCA分析表明,列线图模型的临床价值优于临床模型和放射组学模型。结论列线图模型可有效鉴别AVM脑出血与原发性脑出血,有助于临床决策。

Objective To develop a radiomics nomogram model based on CT to distinguish arteriovenous malformation(AVM)intracerebral hemorrhage from primary intracerebral hemorrhage.Methods One hundred and thirty-five patients with cerebral hemorrhage confirmed by operation in the First Affiliated Hospital of Soochow University were analyzed retrospectively,including 52 patients with AVM cerebral hemorrhage and 83 patients with primary cerebral hemorrhage.Radiomics features were extracted from baseline CT,radiomics score(Radscore)was calculated and radiomic labels were constructed.Multiple logistic regression analysis was used for clinical features combined with CT signs to establish a clinical model.And then the nomogram model was generated according to the Radscore and the clinical model.The ROC curve and decision curve analysis(DCA)were used to evaluate the discrimination performance of the model.Results Six features were selected and used to establish radiomic labels.The clinical model consisted of age(OR:4.739,95%CI 1.382-16.250)and hematoma location(OR:0.111,95%CI 0.032-0.385),while the nomogram model consisted of age,hematoma location and Radscore.In the training group,there was a significant difference between the nomogram model[area under curve(AUC)0.912]and the clinical model(AUC 0.816),the radiomics model(AUC 0.857)(Z=2.776,2.034,P=0.006,0.042,respectively);While in the validation group,there was no significant difference between the nomogram model(AUC 0.919)and the clinical model(AUC 0.788),the radiomics model(AUC 0.810)(Z=1.796,1.788,P=0.073,0.074,respectively).DCA analysis showed that the clinical value of the nomogram model was superior to the clinical model and radiomic model.Conclusion The radiomics nomogram can effectively distinguish AVM-related cerebral hemorrhage from primary cerebral hemorrhage,which is helpful for clinical decision-making.