常规CT征象和直方图参数在肺腺癌不同病理亚型鉴别诊断中的对比研究

Comparative Study of Conventional CT Signs and Histogram Parameters in Differential Diagnosis of Different Pathological Subtypes of Lung Adenocarcinoma

目的比较常规CT征象和直方图参数对腺泡型、非腺泡型肺腺癌的鉴别效能的差异。方法回顾性收集在本院手术的185例患者共195个肺结节,其中腺泡型104个,非腺泡型91个。收集患者的人口学资料、肿瘤标记物。采用单因素分析及二元Logistic回归分析建立常规CT征象模型、直方图(HA)模型及二者联合模型,通过曲线下面积(AUC)比较三种模型对鉴别腺泡型、非腺泡型肺腺癌的鉴别效能。结果单因素分析显示,腺泡型中,纯磨玻璃结节的占比低于非腺泡型(P=0.040);腺泡型中,病灶的CT平均值(P=0.023)、CT中位数数值(P=0.027)、紧凑度(P=0.003)、球形度(P=0.013)、熵(P<0.001)高于非腺泡型,病灶的峰度(P<0.001)、偏度(P<0.001)低于非腺泡型。二元Logistic回归分析以及ROC曲线显示联合模型及HA模型较常规CT征象模型具有更高的鉴别效能(AUC:直方图VS常规CT:0.763 VS 0.550,P<0.001;联合模型VS常规CT:0.768 VS 0.550,P<0.001);联合模型较HA模型无统计学差异(联合模型VS直方图:0.768 VS 0.763,P=0.768)。结论HA参数较常规CT征象能够更有效的鉴别腺泡型和非腺泡型肺腺癌。

Objective To compare the efficacy of conventional CT signs and histogram parameters in differentiating acinar type and non-acinar type lung adenocarcinoma.Methods A total of 195 pulmonary nodules were retrospectively collected from 185 patients accepted operating in our hospital,of which 104 were acinar dominant and 91 were non-acinar dominant.Demographic data and tumor markers were collected.Univariate analysis and binary Logistic regression analysis were used to establish the conventional CT sign model,histogram(HA)model and the combined model.Area under the curve(AUC)was used to compare the efficacy of three models in differentiating acinar type and non-acinar type lung adenocarcinoma.Results Univariate analysis showed that the proportion of pure ground glass nodules in acinar type was lower than that in non-acinar type(P=0.040).The mean CT value(P=0.023),median CT value(P=0.027),compactness(P=0.003),sphericity(P=0.013)and entropy(P<0.001)of acinar lesions were higher than those of non-acinar lesions,while the kurtosis(P<0.001)and skewness(P<0.001)of acinar lesions were lower than those of non-acinar lesions.Binary Logistic regression analysis and ROC curve showed that the combined model and HA model had higher differential efficacy than conventional CT signs.(AUC:histogram VS conventional CT:0.789 VS 0.557,P<0.001;Combined model VS conventional CT:0.802 VS 0.557,P<0.001);There was no statistical difference between the combined model and the HA model(combined model VS histogram:0.802 VS 0.789,P=0.768).Conclusion HA parameters are more effective than conventional CT signs in differentiating acinar and non-acinar lung adenocarcinoma.