摘要
目的:探讨直径≤10 mm的肺结节电子计算机断层扫描(computed tomography,CT)影像学早期影像特征。方法:收集2014年4月至2020年11月经病理证实的肺结节152例,其中良性51例,恶性101例。通过观察肺结节CT图像特征信息,包括结节平均直径、密度(实性、混合密度、磨玻璃密度)、部位、与胸膜的位置关系(胸膜下、非胸膜下)、形态(圆形、卵圆形、不规则形);边界(清楚、模糊)、边缘(分叶及毛刺征)、充气支气管征、空泡征、血管集束征、血管穿行征、胸膜凹陷征及卫星灶,邻近支气管改变。应用深睿全肺增强新冠肺结节AI软件分析,将结果与金标准进行比较,计算敏感度、特异度、准确率、阴性预测值、阳性预测值,寻找其诊断价值的影像学征象。结果:肺结节分析软件诊断结果显示,恶性94例(肺腺癌62例、原位癌30例、类癌1例、鳞癌1例),良性47例(瘤样增生5例、错构瘤3例、炎性肉芽肿34例、炎性假瘤3例、真菌感染结节1例、纤维碳沫结节1例),肺结节分析软件诊断肺结节良恶性灵敏度为93.06%,特异度为92.15%,准确度为92.76%,阳性预测值为95.91%,阴性预测值为87.03%,与手术后病理结果基本一致。结论:基于人工智能肺结节分析软件的CT影像学分析,对肺结节(直径≤10 mm)的定性有重要鉴别价值。
Objective:To investigate the early imaging features of pulmonary nodules which are≤10 mm in diameter by computed tomography(CT).Methods:A total of 152 cases with pulmonary nodules which were pathologically confirmed from April 2014 to November 2020 were collected,including 51 benign cases and 101 malignant cases.Imaging features including average diameter,density(solid,mixed ground glass,purely ground glass),position,relationship between lesions and pleura(subpleural,away from the pleura),shape(round,oval,irregular),boundary(clear,unclear),margin(lobulation,spiculation sign),air-bronchogram sign,vacuole sign,vessel convergence sign,central vessel signs,pleural indentation sign,satellite foci and alteration of the adjacent bronchioles were assessed.The sensitivity,specificity,accuracy,the positive predictive value and negative predictive value were calculated by comparison of the results analyzed by Shenrui Pulmonary Nodule Analysis Software and the gold standard to determine valuable imaging features.Results:According to the analysis of the software,among the 152 cases,there were 94 malignant cases(62 cases of lung adenocarcinoma,30 cases of carcinoma in situ,1 case of carcinoid and 1 case of squamous cell carcinoma),47 benign cases(5 cases of tumor-like hyperplasia,3 cases of hamartoma,34 cases of inflammatory granuloma,3 cases of inflammatory pseudotumor,1 case of fungal infection and 1 case of fibrous carbon nodule).The results of the lung nodule analysis software demonstrated the sensitivity,specificity,accuracy,the positive predictive value and negative predictive value were 93.06%,92.15%,92.76%,95.91%and 87.03%respectively,which were consistent with the pathological results after surgery.Conclusion:CT-imaging analysis based on AI pulmonary nodule analysis software has a great differential value in qualitative diagnosis of pulmonary nodules≤10 mm in diameter.
作者
胡春洪
赖爽
秦正英
邹咏梅
夏钦红
鲁黎
宋雪
吕艳
李信友
尹莉
李咏梅
Hu Chunhong;Lai Shuang;Qin Zhengying;Zou Yongmei;Xia Qinhong;Lu Li;Song Xue;LüYan;Li Xinyou;Yin Li;Li Yongmei(Department of Medical Imaging,Chongqing Jianshe Hospital;Health Examination Center,Chongqing Jianshe Hospital;Nursing Department,Chongqing Jianshe Hospital;Department of Radiology,The First Affiliated Hospital of Chongqing Medical University;Department of Respiratory and Critical Care Medicine,Chongqing Jianshe Hospital)
出处
《重庆医科大学学报》
CAS
CSCD
北大核心
2022年第4期473-478,共6页
Journal of Chongqing Medical University
基金
2019年度重庆市九龙坡区科学技术局资助项目(编号:2019-02-020-Y)。
关键词
肺结节
影像组学
计算机体层摄影术
人工智能
pulmonary nodule
radiomics
computed tomography
artificial intelligence
