逐步判别分析甲状腺良、恶性结节的CT鉴别诊断因素  

Differential diagnosis of benign and malignant thyroid nodules using computed tomography based on stepwise discriminant analysis

作　　者：林少帆[1] 林黛英[1] 吴先衡[1] 杨江爽 郑旭峰[1] 吴耀滨 LIN Shao-fan;LIN Dai-ying;WU Xian-heng(Department of Radiology,Shantou Central Hospital (Shantou Hospital Affiliated to Sun Yat-Sen University),Guangdong 515041,China)

机构地区：[1]汕头市中心医院(中山大学附属汕头医院)放射科,广东515041

出　　处：《放射学实践》2020年第4期473-477,共5页Radiologic Practice

摘　　要：目的:探讨基于逐步判别分析法评估CT检查对甲状腺良、恶性结节的鉴别诊断价值。方法:回顾性分析经本院CT检查及组织学证实的85例125枚甲状腺结节的CT检查资料。其中恶性结节65个,良性结节60个。采用逐步判别分析法对甲状腺结节各相关指标(X1性别,X2年龄,X3结节位置,X4结节形态,X5结节边界,X6结节包膜,X7结节内囊性成分,X8微小钙化,X9增强扫描结节边缘模糊/范围缩小,X10甲状腺边缘中断征,X11淋巴结异常)进行分析,建立判别函数,并采用交互验证法验证判别函数的诊断效能。结果:逐步判别分析法筛选出结节形态、边界、增强扫描结节边缘模糊/范围缩小、甲状腺边缘中断征及淋巴结异常共5个独立预测因素。根据这些因素建立判别函数,甲状腺恶性结节的判别式为Y1=-28.687+6.266X4+6.346X5+8.143X9+16.637X10+8.434X11;甲状腺良性结节的判别式为Y2=-68.726+9.443X4+9.505X5+12.530X9+28.221X10+11.879X11。交互验证法结果显示判别函数对甲状腺良、恶性结节的总体诊断符合率为92%。结论:基于甲状腺良、恶性结节的CT检查采用逐步判别分析法建立的判别函数,对甲状腺结节良、恶性的鉴别诊断准确性较好,可作为一种辅助诊断方法,具有临床实用价值。Objective:To investigate the values of CT examination in the differential diagnosis of benign and malignant thyroid nodules by stepwise discriminant analysis.Methods:The CT data of 85 patients with 125 thyroid nodules(malignant 65,benign 60)were retrospective analyzed.All predictive factors of thyroid nodules were analyzed by stepwise discriminant analysis,including gender(X1),age(X2),location of nodules(X3),shape of nodules(X4),margin of nodules(X5),capsular of nodules(X6),cystic change in nodules(X7),micro-calcification of nodules(X8),enhanced range reduction/obscure(X9),thyroid edge interruption(X10)and abnormal lymph node(X11).The discriminant functions were established,and the diagnostic efficacy was verified by interactive validation.Results:The shape,margin,enhanced range reduction/obscure,thyroid edge interruption,abnormal lymph node were the independent predictive factors according to discriminant analysis.With these 5 factors,the discriminant functions of malignant and benign nodule was established.The function of malignant thyroid nodules was Y1=-28.687+6.266X4+6.346X5+8.143X9+16.637X10+8.434X11;The function of benign thyroid nodules was Y2=-68.726+9.443X4+9.505X5+12.530X9+28.221X10+11.879X11.The results of interactive validation showed that the total accuracy of the discriminant functions in classifying benign and malignant thyroid nodules was 92%.Conclusion:The discriminant functions established by stepwise discriminant analysis based on the CT examination had high accuracy in differential diagnosis of benign and malignant thyroid nodules,thus can be used as an aided method for clinical application.

关 键 词：甲状腺 甲状腺结节 体层摄影术 X线计算机 鉴别诊断 

分 类 号：R814.42[医药卫生—影像医学与核医学] R736.1[医药卫生—放射医学]