预测甲状腺微小癌的多元回归公式  

A Multiple Regression Formula for the Prediction of Thyroid Microcarcinoma

作　　者：张明博[1] 付帅[1] 周昀[1] 吕发勤[1] 唐杰[1] 

机构地区：[1]中国人民解放军总医院超声诊断科,北京100853

出　　处：《中国医学科学院学报》2014年第1期79-85,共7页Acta Academiae Medicinae Sinicae

摘　　要：目的建立预测甲状腺微小癌的多元回归定量公式,确定行超声引导下穿刺活检的界值。方法回顾性分析830个最大径≤1 cm的甲状腺结节的声像图特征及病理结果,评价各声像图特征的诊断价值,采用二分类多元Logistic回归得出预测甲状腺微小癌的定量公式,并评价其诊断价值,得出应用该公式推荐穿刺活检的界值。结果恶性风险较高的声像图特征包括实性(OR=41.97)、微小分叶(OR=25.89)、低回声(OR=10.36)、无声晕(OR=8.38)、边缘不规则(OR=4.26)、纵横比大于1(OR=2.71)、微小钙化(OR=1.92)和粗大钙化(OR=1.28)。多元回归公式诊断甲状腺微小癌的敏感性、特异性和准确性分别为90.9%、54.0%和72.5%。结论多元回归公式可较客观地预测甲状腺微小癌,提供穿刺活检的界值。Objective To establish a quantitative analysis formula for the prediction of thyroid micro- carcinoma and decide the cut-off values for the recommendation of ultrasound-guided biopsy. Methods The ul- trasound characteristics of 830 subcentimeter thyroid nodules were retrospectively analyzed based on pathological results in this study. A diagnostic formula was developed using multivariate binary Logistic regression with the cut- off values for the recommendation of biopsy. The diagnostic values of each feature and the formula were evalua- ted. Results The most suspicious ultrasound characteristics for subcentimeter thyroid nodules were solid echo- structure （ OR = 41.97）, microlobulated margin （ OR = 25.89）, hypoechoic echogenicity （ OR = 10. 36）, no halo （ OR = 8. 38）, irregular margin （ OR = 4. 26）, taller than wide （ OR = 2. 71 ）, microcaleification （ OR = 1.92）, and macrocalcification （ OR = 1.28 ）. The sensitivity, specificity, and accuracy of the formula were 90. 9%, 54. 0%, and 72. 5%, respectively. Conclusion This multiple regression formula is an objective tool for the evaluation of thyroid microcarcinoma, which can provide the cutoff values for the ultrasound guided biopsy.

关 键 词：甲状腺 微小癌 超声 多元回归 

分 类 号：R445.1[医药卫生—影像医学与核医学]