基于CT影像特征构建鉴别肾上腺嗜铬细胞瘤和乏脂性腺瘤的预测模型

Establishment of a prediction model based on CT image features for differential diagnosis of adrenal pheochromocytomas and lipid-poor adenomas

目的 建立基于双期增强CT影像特征预测肾上腺嗜铬细胞瘤(PHEO)和乏脂性腺瘤(Lp-AA)的logistic回归模型,并评估其鉴别价值。方法 收集舟山医院2012年1月至2021年12月行CT增强检查和术后病理检查确诊的PHEO患者70例和LP-AA患者100例,分别记录并对CT影像的连续变量和分类变量进行分析,对差异有统计学意义的因素进行二元logistic回归分析,并建立回归模型。绘制ROC曲线并计算AUC。结果 两组患者肿瘤的大小、CT测量值、增强幅度、增强廓清率、肿瘤的囊变、瘤内血管的差异均有统计学意义(均P<0.05)。二元logistic回归分析结果显示:囊变、增强强化峰值/平扫CT值、平扫CT值是鉴别两者的独立影响因素,模型预测准确度、灵敏度、特异度、阳性预测值、阴性预测值分别为0.847、0.870、0.814、0.870、0.814,用于评价该模型的AUC为0.931。结论 基于双期增强CT影像特征的预测模型对于鉴别PHEO和Lp-AA具有较高的应用价值。

Objective To establish a prediction model based on biphasic contrast-enhanced CT image features in differential diagnosis of adrenal pheochromocytoma(PHEO) and lipid-poor adenoma(Lp-AA). Methods The clinical data and CT findings of 70 patients with PHEO and 100 patients with Lp-AA from January 2012 to December 2021 in Zhoushan Hospital, whose diagnosis was pathologically confirmed, were retrospectively analyzed. CT imaging findings were measured and calculated quantitatively and qualitatively. A differentially diagnostic model was developed and evaluated with receiver operator characteristic(ROC) curve. Results On the quantitative variables of CT findings, there were significant differences in size, CT values, degree of enhancement, and percentage washout of enhancement between two groups(P<0.05). On the qualitative variables, there were significant differences in tralesional necrosis and blood vessels between two groups(P<0.05). Logistic regression analysis showed that necrosis,(the ratio of peak enhanced(PE)/unenhanced CT attenuation values(U) and CTu(unenhanced CT attenuation values) were independent influencing factors in the differential diagnosis. The accuracy, sensitivity, specificity, positive predictive value, negative predictive value of logistic regression model for differential diagnosis between PHEO and Lp-AA were 0.847, 0.870, 0.814, 0.870, 0.814, respectively;the area under the ROC curve(AUC) of the model was 0.931. Conclusion The established prediction model based on biphasic contrast-enhanced CT image features may be used for differential diagnosis between PHEO and Lp-AA.