人工智能肺结节诊断系统对亚实性肺结节良恶性的预测  

Prediction of Benign or Malignant of Subsolid Nodules by Artifiaial Intelligence Pulmonary Nodule Diagnosis System

作　　者：周诚[1,2] 刘文亚[1] ZHOU Cheng;LIU Wenya(The First Affiliated Hospital of Xinjiang Medical University,Urumqi,Xinjiang Province 830000,P.R.China)

机构地区：[1]新疆医科大学第一附属医院,乌鲁木齐830000 [2]新疆医科大学附属肿瘤医院CT室,乌鲁木齐830000

出　　处：《临床放射学杂志》2022年第2期265-268,共4页Journal of Clinical Radiology

摘　　要：目的评价人工智能(AI)肺结节影像诊断系统预测亚实性结节(SN)恶性概率值的效能。方法搜集2017年至2019年经手术及病理结果证实的142个SN,将结节分为原位癌组、浸润性腺癌组、炎性结节组。将患者术前CT检查数据导入AI肺结节诊断系统,记录系统测量的SN的体积、平均CT值、能量值、恶性概率值。比较三组SN在CT平扫、增强扫描动脉期及延迟期中的体积、CT值、能量值及恶性概率值,并对平扫、动脉期及延迟期扫描的各组数据进行配对样本检验。分析根据各期CT图像对各组SN恶性概率值与CT值、体积、能量的相关性。结果共纳入142个SN,其中1组为42个,2组为75个,3组为25个。AI系统对三组SN平扫、动脉期、延迟期所检测恶性概率值分别为1组中9.5(7,84.25)%、89(79.75,90)%、89(68.25,90)%;2组中84(26,89)%、89(85,90)%、89(85,90)%;3组中16(9,79)%、79(19,89)%、70(22,89)%,三组SN在扫描各期的恶性概率值均存在统计学差异(P<0.01)。三组SN在扫描各期的CT值、体积、能量值均存在统计学差异(P<0.001),3组间两两比较均存在统计学差异,每组内平扫与动脉期、平扫与延迟期均存在统计学差异(P<0.001),动脉期与延迟期比较差异均无统计学差异(P>0.05)。CT扫描各期的CT值及体积、能量值对SN的恶性概率值均呈正相关(P均<0.01)。结论 AI肺结节辅助诊断系统提供的一些参数,可以为医师对良恶性结节的鉴别提供一定的帮助。Objective To investigate the efficiency of various parameters on the malignant probability value of benign and malignant nodules by detecting various parameters of pulmonary subsolid nodules(SN) with artificial intelligence(AI) diagnosis system. Methods 142 pulmonary subsolid nodules confirmed by surgical operation and pathology from 2017 to 2019 were collected retrospectively.The nudoles were divided into three groups.Group 1 was carcinoma in situ, and Group 2 was invasive adenocarcinoma.Group 3 was inflammatory nodules.The preoperative CT data of patients were imported into the AI imaging diagnosis system of pulmonary nodules, and the volume, average CT value, energy value and malignant probability value of subsolid nodules measured by the system were recorded.The volume, CT value, energy value and malignant probability of SN in plain and enhanced CT were compared among the three groups.The correlation between malignant probability of SN and CT value, volume and energy was analyzed. Results A total of 142 SN were enrolled, including 42 in group 1,75 in group 2 and 25 in group 3.The malignant probability of SN based on plain CT,arterial phase and delayed phase of enhanced CT was 9.5(7,84.25)%、89(79.75,90)%、89(68.25,90)% in group 1,84(26,89)%、89(85,90)%、89(85,90)% in group 2,16(9,79)%、79(19,89)%、70(22,89)% in group 3,respectively.Statistical differences of nodule volume, CT value, energy value and the malignant probability of 3 phases CT images were found among 3 groups(P<0.001).There were statistically significant differences among the 3 groups, and there were statistically significant differences between plain scan and arterial phase, plain scan and delayed phase in each group( P<0.001),while no statistically difference of malignant probability of SN between arterial phase and delayed phase was found in any group(P>0.05).The CT value, volume and energy of each phase of CT scan were positively correlated with the malignant probability of SN(P<0.01). Conclusion Parameters provided by the AI pulmonary

关 键 词：人工智能 肺结节 肺癌 诊断 

分 类 号：R734.2[医药卫生—肿瘤]