基于CT影像组学构建富血供超小肾癌与乏脂肪血管平滑肌脂肪瘤鉴别诊断的简易预测模型

曹娜; 黄渤琪; 刘志民; 王栋

doi:10.7619/jcmp.20230492

基于CT影像组学构建富血供超小肾癌与乏脂肪血管平滑肌脂肪瘤鉴别诊断的简易预测模型

Establish of a simple predictive model based on CT imaging histology in the differential diagnosis of extra-small renal cell carcinoma with rich blood supply and angiomyolipoma with minimal fat

摘要

摘要:
目的探讨基于CT影像组学构建的简易预测模型对富血供超小肾癌(usRCC)与富血供乏脂肪血管平滑肌脂肪瘤(mfAML)的鉴别诊断价值。
方法收集经术后病理证实为富血供超小肾肿瘤(直径≤2 cm)的71例患者的临床资料, 依据术后病理类型将患者分为usRCC组33例和mfAML组38例。比较2组临床资料、CT影像学表现以及相关CT定量参数，采用二元Logistic回归分析筛选对usRCC与mfAML具有鉴别意义的独立影响因素，并构建基于CT影像组学的简易预测模型，绘制受试者工作特征(ROC)曲线，评价相关CT定量参数和简易预测模型对usRCC与mfAML的鉴别诊断价值。
结果 usRCC组囊变坏死、假包膜征、实质期不均匀强化者占比均高于mfAML组，差异有统计学意义(P < 0.05); usRCC组皮质期CT值、皮质期净强化CT值和实质期净强化CT值均高于mfAML组，差异有统计学意义(P < 0.05)。ROC曲线显示，皮质期CT值、皮质期净强化CT值、实质期净强化CT值鉴别诊断usRCC与mfAML的曲线下面积(AUC)分别为0.702、0.718、0.803。囊变坏死(OR=2.537, 95% CI: 1.125~4.358)、实质期强化均匀性(OR=3.872, 95% CI: 1.327~7.259)、实质期净强化CT值(OR=3.593, 95% CI: 1.290~7.518)均是对usRCC与mfAML具有鉴别意义的独立影响因素(P < 0.05), 基于这3个CT影像组学变量构建简易预测模型。ROC曲线显示, 该模型鉴别诊断usRCC与mfAML的AUC为0.890(95% CI: 0.804~0.976), 敏感度为87.88%, 特异度为76.32%。
结论基于CT影像组学构建的简易预测模型对usRCC与mfAML具有较好的鉴别诊断价值，可为临床制订小肾肿瘤诊疗策略提供重要参考依据。

Abstract:
Objective To explore the differential value of a simple prediction model based on CT imaging histology in the diagnosis of ultra-small renal cell carcinoma (usRCC) with rich blood supply and angiomyolipoma with minimal fat (mfAML).
Methods The clinical data of 71 patients with ultra-small renal tumor (diameter ≤2 cm) with rich blood supply were collected. According to the postoperative pathological types, they were divided into usRCC group (n=33) and mfAML group (n=38). Clinical data, CT imaging manifestations, and related CT quantitative parameters were compared, independent influencing factors with differential significance for usRCC and mfAML were screened using binary Logistic regression, and a simple predictive model based on CT imaging histology was constructed. Receiver operating characteristic (ROC) curves were drawn to evaluate differential value of relevant CT quantitative parameters and predictive model for usRCC and mfAML.
Results The proportion of cystic necrosis, pseudocapsule sign and parenchymal phase heterogeneous enhancement in the usRCC group was higher than that in the mfAML group (P < 0.05). The CT value of cortical phase, enhanced CT value of cortical phase and parenchymal phase in the usRCC group were also significantly higher than those in the mfAML group (P < 0.05). The areas under the curve (AUCs) of differential diagnosis of usRCC and mfAML by CT value of cortical phase, enhanced CT value of cortical phase and parenchymal phase were 0.702, 0.718 and 0.803, respectively. Cystic necrosis (OR=2.537; 95% CI, 1.125 to 4.358), parenchymal enhancement uniformity (OR=3.872; 95% CI, 1.327 to 7.259), and parenchymal net enhancement CT value (OR=3.593; 95% CI, 1.290 to 7.518) were independent influencing factors for differentiated diagnosis of usRCC and mfAML (P < 0.05), thus a simple prediction model was constructed based on the three CT image omics variables. The ROC curve showed that the AUC of the model for differential diagnosis of usRCC and mfAML was 0.890 (95% CI, 0.804 to 0.976), the sensitivity was 87.888, and the specificity was 76.32%.
Conclusion The simple prediction model based on CT imaging histology has a good value in differential diagnosis of usRCC and mfAML, and provides an important reference for clinical diagnosis and treatment of small renal tumors.

HTML全文

参考文献(22)

施引文献

资源附件(0)