孤立性肺小结节良恶性判别的初步定量研究

Preliminary Quantitative Study on the Correlation Between Solitary Pulmonary Nodules and Pathology

目的定量分析孤立性肺小结节(≤2 cm)影像征象、临床资料与病理的关系,构建并验证相关多重线性回归方程模型,将该模型与国内李运模型、杨娟模型、国外Mayo模型以及VA模型进行比较。方法回顾性分析经手术病理证实的107例(A组)孤立性肺小结节CT征象、患者年龄及性别,根据病理结果分为良、恶性组,建立恶性组、良性组及总体组的多重线性回归方程模型,预测结节病理直径并判别良恶性,另外搜集65例(B组)资料验证本文模型并与4个模型进行比较。结果方程模型分别为:Y(总)=0.156-0.005×X1+0.007×X2+0.826×X3+0.214×X4-0.019×X5-0.007×X6-0.019×X7-0.087×X8-0.017×X9+0.025×X10。Y(恶)=0.235-0.005×X1-0.019×X2+0.821×X3+0.133×X4-0.017×X5+0.064×X6-0.014×X7-0.089×X8-0.029×X9;Y(良)=0.168-0.002×X1-0.026×X2+0.797×X3+0.216×X4-0.1×X5-0.109×X6-0.007×X7-0.088×X8+0.134×X9-0.064×X10。良、恶性病例中肺小结节实际病理直径的95%可信区间分别为(1.160,1.407)、(1.040,1.268)。总体模型、恶性模型、良性模型、实际病理结果及验证组的模型预测曲线下面积(AUC)分别为0.684±0.083、0.730±0.078、0.645±0.084、0.699±0.07、0.673±0.077。良性模型检验结果变量无统计学意义(P>0.05),而总体模型、恶性模型、实际病理结果及模型预测的检验结果变量均有统计学意义(P<0.05)。利用B组数据验证本文模型的AUC为0.673±0.077,李运模型的AUC为0.787±0.062,杨娟模型的AUC为0.589±0.078,Mayo模型的AUC为0.723±0.067,VA模型的AUC为0.638±0.075。本研究模型的AUC值与4个模型的AUC值差异存在统计学意义(P<0.05)。结论肺小结节的影像测量直径、密度、分叶征、毛刺征、空泡征或支气管充气征、胸膜凹陷征、血管集束征、钙化征及患者年龄、性别是数学预测模型方程判断其良、恶性的独立相关因素,其中恶性结节数学预测模型的准确性较高,总体模型次之,本研究模型比国内李运模型、杨娟模型、国外Mayo模型以及VA模型在影像征象方面更有优势,具有较好的临床应用价值。

Objective To quantitatively analyze the relationship between imaging signs,clinical data and pathology of solitary pulmonary nodules(≤2 cm),and to construct and validate the relevant multiple linear regression equation model. The model is related to the domestic Li Yun model,Yang Juan model,Foreign MAYO models and VA models are compared. Methods A retrospective analysis of 107 cases(group A) of isolated pulmonary nodules with CT signs,age and gender of patients with pathological findings were divided into good and malignant groups,and malignant,benign and general groups were established. Multiple linear regression equation models were used to predict the pathological diameter of nodules and to distinguish between benign and malignant. Another 65 cases(group B) were collected to verify the model and compared with the four models. Results The equation models are:Y(total)=0.156-0.005×X1+0.007×X2+0.826×X3+0.214×X4-0.019×X5-0.007×X6-0.019×X7-0.087×X8-0.017×X9+0.025×X10. Y(evil)=0.235-0.005×X1-0.019×X2+0.821×X3+0.133×X4-0.017×X5+0.064×X6-0.014×X7-0.089×X8-0.029×X9;Y(liang)=0.168-0.002×X1-0.026×X2+0.797×X3+0.216×X4-0.1×X5-0.109×X6-0.007×X7-0.088×X8+0.134×X9-0.064×X10.The 95% confidence interval for the actual pathological diameter of pulmonary nodules in benign and malignant cases was(1.160,1.407),(1.040,1.268). The area under the model prediction curve(AUC) of the overall model,the malignant model,the benign model,the actual pathological results,and the validation group were 0.684±0.083,0.730±0.078,0.645±0.084,0.699±0.07,and 0.673±0.077,respectively. The outcome variables of the benign model test were not statistically significant(P>0.05),but the overall model,the malignant model,the actual pathological results and the test results predicted by the model were statistically significant(P<0.05). Using the data of Group B,the AUC of the model was 0.673±0.077,the AUC of the Li Yun model was 0.787±0.062,the AUC of the Yang Juan model was 0.589±0.078,the AUC of the Mayo model was 0.723±0.067,and the AUC of the VA model was 0.638±. 0.075. The AUC values of the model in this paper were statistically different from the AUC values of the four models(P<0.05). Conclusion The image of the small pulmonary nodules is measured by diameter,density,lobulation sign,burr sign,vacuole sign or bronchial aeration sign,pleural indentation sign,vascular bundle sign,calcification sign and patient age and gender. The independent and malignant independent related factors,the mathematical prediction model of malignant nodules is more accurate,and the overall model is second. The model of this paper is more than the domestic Li Yun model,Yang Juan model,foreign MAYO model and VA model in terms of image signs and has good clinical application value.