摘要
目的:研究人工智能能否替代CT影像诊断医生的巨大工作量,且能为临床医生在肺结节早期诊断、良恶性判断、治疗策略及手术方式选择等方面提供重要的指导意义。方法:通过128层螺旋CT胸部低剂量扫描方式,进行多面重组(MPR)、容积重现(VR)等三维重建后处理,对比分析人工阅片检出肺结节的敏感度、符合率及消耗时间;运用σ-DiscoverLung智能肺结节检测系统检出肺结节、自动分割病灶、自动测量参数、自动分析良恶性、自动结构式报告等的敏感度、符合率及消耗时间;运用统计学方法对比分析人工阅片与σ-DiscoverLung智能肺结节检测系统检测肺结节的敏感度、符合率、消耗时间及优劣势。结果:项目组在研究中发现σ-DiscoverLung半分钟内就能够完成所有肺结节的诊断,经过对比分析、统计,其对肺结节诊断的敏感性达到95.6%,良恶性判断的准确率达到89.5%,假阳性率25.5%。人工阅片与AI阅片相互补充、相互映证,提高了肺癌早期诊断的符合率,为临床在肺结节早期诊断、良恶性判断、治疗策略及手术方式选择等方面提供重要的指导意义。结论:人工阅片与σ-DiscoverLung智能阅片可以相互补充、相互映证,从而提高肺癌早期诊断的符合率,指导临床医生治疗方案的选择。
Objective To study whether artificial intelligence can replace the huge workload of CT imaging diagnostic doctors,and provide important guidance for clinicians in early diagnosis of pulmonary nodules,benign and malignant judgment,treatment strategies and surgical options.Methods By means of 128-slice spiral CT low-dose thoracic scanning,three-dimensional reconstruction post-processing such as MPR and VR were carried out,and the sensitivity,coincidence rate and consumption time of detecting pulmonary nodules by manual reading were compared and analyzed;pulmonary nodules were detected byσ-Discover Lung intelligent pulmonary nodule detection system,automatic segmentation of lesions,automatic measurement parameters and automatic analysis were performed.Sensitivity.coincidence rate and consuming time of malignant and automatic structured reports were compared and analyzed by statistical methods.Sensitivity,coincidence rate,consuming time,advantages and disadvantages of manual film reading andσ-Discover Lung intelligent pulmonary nodule detection system were compared and analyzed.Results The project team found thatσ-Discover Lung could complete the diagnosis of all pulmonary nodules in half a minute.Through comparative analysis and statistics,the sensitivity ofσ-Discover Lung to the diagnosis of pulmonary nodules reached 95.6%,the accuracy of benign and malignant judgment reached 89.5%,and the false positive rate was 25.5%Artificial film reading and AI film reading complement each other and reflect each other’s evidence,which improves the coincidence rate of early diagnosis of lung cancer,and provides important guidance for clinical diagnosis of pulmonary nodules,benign and malignant judgment,treatment strategy and choice of surgical methods.Conclusion Artificial film reading andσ-Discover Lung intelligent film reading can complement and reflect each other,thus improving the coincidence rate of early diagnosis of lung cancer and guiding the choice of treatment options for clinicians.
作者
王杜春
任龙
刘宁川
杨柳
何杰
Wang Duchun;Ren Long;Liu Ningchuan;Yang Liu;He Jie(Department of Radiology,People's Hospital of Gaoping District,Nanchong,Sichuan 637100,China)
出处
《影像研究与医学应用》
2019年第16期39-41,共3页
Journal of Imaging Research and Medical Applications
基金
南充市科技局基金(18YFZJ0007)~~
