基于CT形态学的逐步判别分析在甲状腺结节中的鉴别

Stepwise Discriminant Analysis Based on CT Morphology in the Differentiation of Thyroid Nodule

目的探讨基于CT形态学的逐步判别函数在甲状腺结节中的鉴别价值。方法分析90名患者114例甲状腺结节,经组织学确认67例良性结节,47例恶性结节。对甲状腺结节CT特征通过单变量统计和逐步判别分析其与病理组织学结果符合度。结果影像报告与病理无显著差异(P>0.05),灵敏度85.11%,特异度86.57%,正确度85.97%。形态欠规则、边界欠清、淋巴结异常、微钙化、“咬饼”征及增强后结节边界缩小或模糊常见于恶性结节(P均<0.05),囊性为主、包膜完整以及高强化多见于良性病变(P均<0.05);将有统计学意义的CT特征纳入逐步判别分析,共筛出病灶边界、包膜、囊变为主、“咬饼”征、强化程度、增强后边界缩小或模糊及淋巴结异常7个独立预测指标,根据这些指标建立判别函数,回顾性判别甲状腺良、恶性结节诊断符合率达93.9%。结论基于CT形态学的逐步判别函数可以显著提升术前诊断甲状腺良、恶性结节的准确率。

Objective To evaluate the differential diagnostic performance of the stepwise discriminant function based on CT morphology in thyroid nodules.Methods This study retrospectively enrolled 90 patients with 114 thyroid nodules(malignant:benign=47:67).The univariate statistics and stepwise discriminant analysis were used to analyze the conformity between CT morphology and histopathological results of thyroid nodules.Results There was no obvious difference between imaging findings and pathological results(P>0.05).The Sensitivity,specificity and accuracy were 85.11%,86.57%,and 85.97%,respectively.The malignant nodules are characterized by irregular shape,unclear boundary,abnormal lymph nodes,microcalcification,cake bite sign and boundary reduction/blurring after enhancement(P<0.05).While the cystic predominance,intact capsule and high enhancement were more common in benign lesions(P<0.05).A total of 7 independent predictive indexes were selected by stepwise discriminant analysis,including lesion boundary,capsule,cystic predominance,cake bite sign,enhancement degree,boundary reduction/blurring after enhancement and abnormal lymph nodes.By means of eatablishing discriminant function based on these indexes,the accuracy of retrospectively discriminating benign and malignant thyroid nodule was 93.9%.Conclusion The accuracy of diagnosis of benign and malignant thyroid nodules could be significantly improved before surgery by the stepwise discriminant function based on CT morphology.