计算机辅助系统联合超声对甲状腺结节的鉴别诊断价值

The differentiation value of computer-aided diagnosis system in ultrasound diagnosis of thyroid nodule

目的研究计算机辅助诊断(computer-aided diagnosis,CADx)系统联合超声对甲状腺结节良恶性的鉴别诊断效果。方法对2017年10月—2018年4月超声科检查的已有病理结果证实的甲状腺结节98例患者,资历相同的两组医师分别在有无CADx辅助情况下,对图像进行判读,按ACR指南进行TI-RADS分级,以病理结果为金标准,比较两组的判读能力。结果人工判读组和有CADx辅助组对98个甲状腺结节进行分级的ROC曲线下面积分别为0.8025、0.9024。当以TI-RADS V级作为诊断恶性标准时,人工判读组和CADx辅助组的灵敏度为78.13%、92.19%(P<0.05),差异有统计学意义;特异度为82.35%、88.24%(P>0.05),差异无统计学意义。CADx系统中特征参数低回声、有强回声光点、纵横比>1、无囊性回声均有效鉴别甲状腺良恶性结节,差异有统计学意义(P<0.05)。结论 CADx系统能够有效鉴别甲状腺的良恶性结节,提高超声医师对甲状腺结节的诊断效果。

Objective To study the ultrasound diagnostic efficacy of computer-aided diagnosis(CADx)system for differentiating benign from malignant thyroid nodules.Methods The 98 patients with pathologically confirmed thyroid nodules after confirmed by sonography from October 2017 to April 2018 were enrolled in this study,and the two groups of doctors with same professional titles interpreted the images with or without CADx assistance,ACR-guided took classification;the pathological results were as the gold standard,for comparing the interpretation ability of the two groups.Results The areas under the ROC curve for classification of 98 thyroid nodules in manual interpretation group and CADx-assisted group were 0.8025 and 0.9024,respectively.When TI-RADS V grade was used as a diagnostic criterion,the sensitivities of manual interpretation group and CADx-assisted group were 78.13%and 92.19%(P<0.05)respectively,which means the difference was statistically significant;in the other hand,the specificities respectively were 82.35%and 88.24%(P>0.05),i.e the difference was not statistically significant.In the CADx system,the characteristic parameters of low echo,strong echogenic spots,aspect ratio greater than 1,and no cystic echo were all valid for distinguishing benign and malignant thyroid nodules,i.e there were statistically significant differences(P<0.05).Conclusion The CADx system can effectively differentiate benign ones from malignant thyroid nodules,and improve the diagnostic efficiency of the sonographers for thyroid nodules.