摘要
目的探讨基于人工智能(AI)技术联合CT平扫影像(NCCT)特征对脑内早期血肿扩大(EHE)的预测价值。方法搜集2018年6月—2021年6月136例自发性脑出血患者的临床和头颅NCCT资料进行回顾性研究。基于AI技术精确分割脑内血肿,测量血肿体积,判断是否存在EHE。采用χ^(2)检验与二元Logistic回归分析岛征、卫星征、混合征、漩涡征、黑洞征、液平征及其联合征象预测EHE的价值,χ^(2)检验和t检验分析临床资料与EHE的相关性。结果头颅NCCT图像显示黑洞征(OR=11.250,95%CI=4.453~28.423,P<0.001)、卫星征(OR=5.779,95%CI=2.322~14.384,P<0.001)、岛征(OR=4.882,95%CI=2.116~11.265,P<0.001)与漩涡征(OR=2.676,95%CI=1.222~5.857,P=0.014)对预测EHE均有统计学意义,是EHE的危险因素。联合黑洞征或卫星征预测EHE的准确率78.7%、特异度82.0%、阴性预测值88.2%、曲线下面积(AUC)0.757,优于单一影像征象及与其他征象的组合。临床资料中,年龄(t=-2.120,P=0.036)、急诊格拉斯哥昏迷评分(GCS)(t=2.763,P=0.007)、出院改良Rankin评分(mRS)(t=-2.992,P=0.003)、高血压史(χ^(2)=4.925,P=0.026)、冠心病史(χ^(2)=4.089,P=0.043)及肾脏病史(χ^(2)=6.543,P=0.011),均与EHE相关性显著。结论基于AI技术,头颅NCCT检查结果显示的黑洞征或卫星征联合应用对EHE有较高的预测价值。
Objective To explore the predictive value of early hematoma expansion(EHE)based on artificial intelligence(AI)technology combined with noncontrast computed tomographic(NCCT)image features.Methods A retrospective study was conducted to collect the clinical and skull NCCT scan image data of 136 patients with spontaneous cerebral hemorrhage from June 2018 to June2021 through the inclusion criteria and exclusion criteria.AI technology was used to segment intracerebral hematoma accurately and measured hematoma volume to determine the presence of EHE.The value of the island sign,satellite sign,mixed sign,blend sign,black hole sign,fluid level,and their combined signs in predicting the EHE were analyzed usingχ^(2)test and binary logistic regression.The correlation between clinical data and the EHE was analyzed byχ^(2)test and t test.Results Black hole sign(OR=11.250,95%CI=4.453-28.423,P<0.001),satellite sign(OR=5.779,95%CI=2.322-14.384,P<0.001),island sign(OR=4.882,95%CI=2.116-11.265,P<0.001)and swirl sign(OR=2.676,95%CI=1.222-5.857,P=0.014)were both statistically significant for predicting the EHE and were risk factors for the EHE.The combined black hole sign or satellite sign predicted the EHE with78.7%accuracy,82.0%specificity,88.2%negative predictive value,and 0.757 area under the curve(AUC),which was superior to single imaging signs and combinations with other signs.Of the clinical data,age(t=-2.120,P=0.036),emergency Glasgow Coma Score(GCS)(t=2.763,P=0.007),discharge modified Rankin Score(mRS)(t=-2.992,P=0.003),history of hypertension(χ^(2)=4.925,P=0.026),history of coronary heart disease(χ^(2)=4.089,P=0.043)and history of kidney disease(χ^(2)=6.543,P=0.011)were significantly associated with the EHE.Conclusion Based on AI technology,the skull NCCT test showed that combining a black hole sign or satellite sign with high predictive value for EHE.
作者
陈越
陈勇
严梓伊
范妍
潘韵涵
周丽芳
丁建平
戚乐
CHEN Yue;CHEN Yong;YAN Ziyi;FAN Yan;PAN Yunhan;ZHOU Lifang;DING Jianping;QI Le(School of Basic Medicine,Hangzhou Normal University,Hangzhou 311121,China;School of Clinical Medicine,Hangzhou Normal University,Hangzhou 310015,China;Department of Radiology,Affiliated Hospital of Hangzhou Normal University,Hangzhou 310015,China)
出处
《中国研究型医院》
2022年第6期41-46,共6页
Chinese Research Hospitals
基金
杭州市生物医药和健康产业发展扶持科技专项(2021WJCY111)
杭州师范大学“本科生创新能力提升工程”项目。
关键词
脑出血
血肿
体层摄影术
X线计算机
人工智能
影像征象
Cerebral hemorrhage
Hematoma
Tomography
x-ray computed
Artificial intelligence
Image markers