目的：探索一种可以较为客观、准确、实用的甲状腺结节恶性风险系数预测法。方法：2013年5月—2016年2月在六安市中医院外科系统的所有被检测出甲状腺结节并住院手术的305例患者的417个结节，以预实验选定的11个超声征象作为自变量，以结节病理性质作为因变量，采用Logistic二分类回归分析得出回归常数项和偏回归系数，经检验后，带入公式   P=en（1+en），n=β0+β1X1+......+βpXp，得出单个结节的恶性风险系数预测公式。计算出本次试验所有结节的恶性风险系数，随访每一例实验病例病理报告，绘制ROC曲线，并确定cut-off值。将恶性风险系数预测公式和cut-off值应用于临床64例甲状腺结节患者的超声诊断并随访病理报告，计算正确率、灵敏度、特异度、阳性预测值、阴性预测值等指标。结果：甲状腺结节恶性风险系数预测公式预测甲状腺结节性质的部分指标如下：正确率89.1%；灵敏度91.7%；特异度88.5%；Youden指数80.2%等。结论：甲状腺结节恶性风险系数预测公式是一种能较为易行、准确、客观和可量化的甲状腺结节恶性风险评估工具。

Objective: To explore a more objective, accurate and practical method for predicting the malignant risk coefficient of thyroid nodules. Method: From May 2013 to February 2016, 417 nodules were detected in 305 patients with thyroid nodules in surgical system at the Lu’an Hospital of Traditional Chinese Medicine. In the preoperative ultrasound examination, the information of 11 ultrasonic signs were recorded according to the pilot experiment. Benign and malignant as the dependent variable, and 11 ultrasound signs as the independent variables, regression coefficient and partial regression coefficients are obtained by Logistic regression analysis. After examination, the formula P=en(1+en), n=β0+β1X1+......+βpXp, was adopted and a formula for predicting the malignant risk coefficient of a single nodule became available to calculate the risk coefficient of each case. Follow up was performed in each case, the ROC curve was drawn, and the cut-off value was determined. The prediction formula for the malignant risk coefficient and cut-off values were applied to the ultrasonic diagnosis of 64 patients with thyroid nodules. The accuracy, sensitivity, specificity, positive predictive value and negative predictive value were calculated. Result: Basing on the formula of risk factor prediction for thyroid nodules, the correct rate was 89.1%, sensitivity was 91.7%, specificity was 88.5%, and Youden index was 80.2%. Conclusion: The formula for predicting the malignant risk of thyroid nodules by ultrasound is a relatively easy, accurate, objective and quantifiable assessment tool.