摘要
目的:分析人工智能辅助CT密度定量对良恶性肺结节的诊断价值。方法:回顾分析2019年1月至2022年5月三明市第二医院、永安市立医院经穿刺活检或手术病理证实的52例患者的临床及影像资料,将原始数据由PACS传至人工智能肺结节辅助软件进行分析。采用两组间t检验、工作特征曲线,评估人工智能(artificial intelligence,AI)定量参数对良恶性肺结节病灶的预测价值。结果:人工智能辅助下肺腺癌患者与良性肺结节患者的定量指标:结节长径和短径、CT最大值、CT最小值、结节体积无统计学意义(P>0.05),CT平均值、CT值标准差、实性占比存在统计学意义(P<0.05)。ROC曲线下面积,CT值标准差(AUC=0.754)、实性占比(AUC=0.743)、CT值平均(AUC=0.735),CT平均值联合CT标准差(AUC=0.767)、CT值标准差联合实性占比(AUC=0.766)可以提高预测价值。结论:人工智能辅助下的肺结节CT密度定量参数,对鉴别肺腺癌与良性肺结节具有诊断意义,CT平均值联合CT标准差的预测价值最高。
Objective:To analyze the diagnostic value of artificial intelligence assisted CT densitometry in benign and malignant pulmonary nodules.Methods:The clinical and imaging data of 52 patients confirmed by puncture biopsy or surgical pathology in Sanming second hospital and Yong’an municipal hospital from January 2019 to may 2022 were retrospectively analyzed,The original data were transmitted from PACS to the artificial intelligence pulmonary nodule assistant software for analysis.The predictive value of quantitative parameters of artificial intelligence(AI)for benign and malignant pulmonary nodules was evaluated by t-test and working characteristic curve between two groups.Results:the quantitative indexes of patients with lung adenocarcinoma and benign pulmonary nodules assisted by artificial intelligence:There was no significant difference in the long and short diameters of the nodules,the maximum value of CT,the minimum value of CT and the volume of the nodules(P>0.05),The mean value of CT,the standard deviation of CT value and the proportion of solid matter were statistically significant(P<0.05).The area under ROC curve,the standard deviation of CT value(AUC=0.754),the proportion of solid matter(AUC=0.743),the average of CT value(AUC=0.735),The combination of CT mean value with CT standard deviation(AUC=0.767)and CT standard deviation with the real proportion(AUC=0.766)can improve the predictive value.Conclusion:the quantitative parameters of CT density of pulmonary nodules assisted by artificial intelligence have diagnostic significance in differentiating lung adenocarcinoma from benign pulmonary nodules.The predictive value of CT mean value combined with CT standard deviation is the highest.
作者
傅良飞
张建平
孙微
Fu Liangfei;Zhang Jianping;Sun Wei(Department of Medical Imaging,Yongan Municipal Hospital,Yongan,Fujian 366000;Department of Medical Imaging,The Second Hospital of Sanming,Sanming,Fujian 366000)
出处
《现代医用影像学》
2022年第12期2179-2183,2192,共6页
Modern Medical Imageology
关键词
人工智能
CT密度定量
肺结节
良恶性
artificial intelligence
CT density quantification
pulmonary nodules
benign and malignant