基于能谱CT及超声初步建立的Logistic回归模型对甲状腺良恶结节鉴别诊断价值的对比研究

Comparative Study on the Differential Diagnosis Value of Logistic Regression Model Based on Spectral CT and Ultrasonography for Benign and Evil Thyroid Nodules

目的探讨基于能谱CT表现、超声表现分别建立的Logistic回归模型对甲状腺良恶性结节的鉴别诊断价值。方法回顾性分析我院2020年3月~2022年3月手术治疗的110例甲状腺结节患者的临床资料,整理并总结术前能谱CT及超声表现。二元Logistic逐步回归法分别建立术前能谱CT及超声表现鉴别诊断甲状腺结节良恶性的Logistic回归模型,并绘制ROC曲线,对比分析各自拟合的联合预测因子对其的鉴别诊断效能。结果术后病理显示,恶性结节患者32例,共36个结节,其中甲状腺乳头状癌30例,滤泡状癌1例,未分化癌1例;良性结节患者78例,共98个结节,其中结节性甲状腺肿66个,甲状腺腺瘤32个。能谱CT表现为有边缘截断征,无包膜,较低的动脉期γ_(HU)、平扫IC及动脉期NIC是恶性结节的独立危险因素(P<0.05),超声表现为形态不规则、钙化、无声晕或欠完整、较高血流分布分级是恶性结节的独立危险因素(P<0.05)。能谱CT及超声表现模型拟合度均较好,有良好的校准能力(P>0.05),能谱CT模型对甲状腺良恶性结节的鉴别诊断价值明显优于超声模型(AUC:0.943 vs.0.859)(P<0.05)。结论与超声相比,以能谱CT表现建立的Logistic回归预测模型鉴别诊断甲状腺结节良恶性的临床价值更高,更具科学性,更有助于为临床治疗方案的选择提供参考,而对于基层医院,术前超声预测模型也是一种较可靠的鉴别诊断手段。

Objective To investigate the value of Logistic regression model based on energy spectrum CT and ultrasound findings in differential diagnosis of benign and malignant thyroid nodules.Methods To retrospectively analyze the clinical data of 110 patients with thyroid nodules undergoing surgical treatment in our hospital from March 2020 to March 2022,to organize and summarize the preoperative spectral CT and ultrasonography manifestations of benign and malignant thyroid nodules.Binary Logistic stepwise regression was used to establish the Logistic regression models for the differential diagnosis of benign and malignant thyroid nodules by energy spectrum CT and ultrasound before surgery,and ROC curve was drawn.To compare and analyze the differential diagnostic efficacy of the combined predictors.Results Postoperative pathology showed 32 patients with 36 malignant nodules,including 30 cases of papillary thyroid carcinoma,1 case of follicular carcinoma and 1 case of undifferentiated carcinoma.78 patients with benign nodules,a total of 98 nodules,66 nodules were nodular goiter and 32 nodules were thyroid adenoma.Spectral CT showed marginal truncation sign,no capsule,low arterial phaseγ_(HU),IC and NIC were independent risk factors for malignant nodules(P<0.05).Ultrasonography findings of irregular shape,calcification,silent halo or incomplete,high blood flow distribution grade were independent risk factors for malignant nodules(P<0.05).The spectral CT and ultrasonography models had good fit and good calibration ability(P>0.05).The spectral CT model was significantly better than the ultrasonography model in the differential diagnosis of benign and malignant thyroid nodules(AUC:0.943vs.0.859),and the difference had statistical significance(P<0.05).Conclusion Compared with ultrasonography,Logistic regression prediction model based on energy spectrum CT findings has higher clinical value in the differential diagnosis of benign and malignant thyroid nodules,which is more scientific and helpful to provide a reference for the selection of clinical treatment.For primary hospitals,the preoperative ultrasonic prediction model is also a reliable means of differential diagnosis.