基于“人工智能”的CT定量分析对肺部结节的临床应用价值

Quantitative CT Analysis of Pulmonary Nodules based on Artificial Intelligence

目的探讨基于"人工智能"的CT定量分析对肺部结节的临床应用价值。方法选取70例2019年1月至2019年12月我院收治并确诊的肺部结节病变患者为研究对象,采用Siemens双源CT对患者进行平行扫描及增强扫描,根据CT影像数据观察分析患者肺部结节病灶检出情况(敏感度、特异度)、结节影像学特征、病灶强化情况及肺部结节谱曲线斜率(K)、病灶钙含量(Ca)、标准化碘浓度(NIC)、有效原子序数(Eff-Z)值变化。结果CT扫描技术诊断肺部结节良、恶性敏感度为96.88%,特异度为97.37%,准确率97.13%;恶性结节和良性结节肺部结节圆形及类圆形、形态不规则、磨玻璃结节、血管集束征、空泡征、分叶征及胸膜凹陷征数比较有统计学差异(P<0.05);鳞状细胞癌组织中央可见空腔和细胞坏死,肺细胞癌影像图边界清晰,伴分叶征,腺癌影像图表现为分叶状或圆形的孤立结节,且边缘呈毛刺状;结节病灶矢状位CT平扫图显示,右肺上尖叶存在小块清晰均匀阴影,增强扫描结果显示肺部病灶轻、中度强化;恶性结节组K、NIC值明显高于良性结节组,有统计学差异(P<0.05);恶性结节组Ca、Eff-Z值明显低于良性结节组,有统计学差异(P<0.05)。结论基于"人工智能"的CT定量分析对肺部结节良、恶性诊断准确率较高,且可进一步鉴别肺内小病灶,对肺部结节患者具有重要临床意义和应用价值。

Objective To investigate the clinical value of quantitative CT analysis of pulmonary nodules based on artificial intelligence.Methods 70 patients with pulmonary nodules treated in our hospital between January 2019 and December 2019 were selected,and all patients received the Siemens dualsource CT scans.According to CT image data,the detection of pulmonary nodule lesions(sensitivity,specificity),imaging characteristics of pulmonary nodules,lesion enhancement,the slope of pulmonary nodule curve(K),lesion calcium content(Ca),normalized iodine concentration(NIC),and effective atomic number(Eff-Z)were observed and analyzed.Results The sensitivity,specificity,and accuracy of CT scanning in the diagnosis of benign and malignant pulmonary nodules were 96.88%,97.37%,and 97.13%,respectively.The CT image features including round shape,quasi-round shape,irregular shape,ground-glass opacity,vessel convergence sign,vacuole sign,lobulation sign and pleural indentation sign between benign and malignant pulmonary nodules(P<0.05).In squamous cell carcinoma,there was space and cell necrosis in the center,the images of lung cell carcinoma showed clear border with lobulation,and images of adenocarcinoma showed a lobulated or round solita ry nodules with burrs on edge.Sagittal CT plain scan showed small clear and uniform shadows in the upper apical lobe of the right lung.The enhanced scan showed mild and moderate enhancement of the pulmonary lesions.The K and NIC values in the malignant nodules group were significantly higher than those in the benign nodules group(P<0.05).The Ca and Eff-Z values in the malignant nodules group were significantly lower than those in the benign nodules group(P<0.05).Conclusion Quantitative CT analysis based on"artificial intelligence"has a high diagnostic accuracy rate for benign and malignant pulmonary nodules.It can further identify small lesions in the lungs,which has important clinical significance and application value for patients with pulmona ry nodules.