CT影像组学模型预测CNLCⅠ~Ⅱ期肝细胞癌切除术后早期复发的研究

Radiomics Model Based on CT Predicts Early Recurrence after Surgical Resection of CNLC Stage I-I Hepatocellular Carcinoma

目的建立术前CT影像组学预测模型对中国肝癌分期(CNLC)Ⅰ~Ⅱ期肝细胞癌(HCC)切除术后早期复发进行预测。方法回顾性分析接受手术切除的CNLCⅠ~Ⅱ期HCC患者153例的资料。用3D slicer软件勾画肿瘤感兴趣区(ROI),用pyradiomics包提取影像组学特征。基于LASSO算法进行特征筛选、并建立影像组学标签(Rad-score)。采用单因素Logistic回归和多因素Logistic逐步回归法确立独立预测因子,构建影像组学预测模型和临床预测模型。用受试者工作特征曲线(ROC)曲线下面积(AUC)来比较模型的区分度,用校准曲线评估模型的校准度,用临床决策曲线分析(DCA)评估模型的临床应用价值。结果Rad-score、瘤内供血动脉、肝功能白蛋白-胆红素分级(ALBI分级)、性别是独立预测因子。影像组学模型具有良好的预测效能(AUC:训练组0.900,验证组0.853),优于临床模型(AUC:训练组0.823,验证组0.741)。校准曲线显示影像组学模型具有良好的校准度。DCA显示阈值概率在0.1~1.0时,影像组学模型的净获益要高于临床模型。结论基于CT影像组学的预测模型对CNLCⅠ~Ⅱ期HCC患者手术切除后早期复发具有良好的预测能力,可为临床医师筛选高复发风险人群、制定个体化诊疗方案提供参考信息。

Objective To establish a preoperative radiomics model based on CT to predict the early recurrence after resection of China Liver Cancer staging(CNLC)Ⅰ-Ⅱhepatocellular carcinoma(HCC).Methods This study retrospectively analyzed 153 CNLCⅠ-ⅡHCC patients who underwent radical resection.The regions of interest were delineated with 3D slicer.The radiomics features were calculated using the pyradiomics package.The least absolute shrinkage and selection operator(LASSO)analysis were performed for Rad-score establishment.Univariate logistic regression and multivariate logistic regression stepwise method were sequentially performed to establish a radiomics model and a clinical model.The area under the receiver operating characteristic curve(AUC)was used to compare the differentiation of the models,the calibration curve was used to evaluate the calibration of the models,and the decision curve analysis(DCA)was used to evaluate the clinical application value of the models.Results The independent predictors were Rad-score,albumin-bilirubin(ALBI)grade,intratumoral arteries and sex.The radiomics model achieved an AUC of 0.900/0.853 in the training/validation set,which was higher than the clinical model(0.832/0.741 in the training/validation set).The calibration curves show the radiomics model with good calibration.The DCA analysis showed that when the threshold probability was around 0.1 to 1.0,the net benefit of radiomics model was higher than that of the clinical model.Conclusion The preoperative radiomics nomogram is helpful in assessing early recurrence after resection of HCC and may be beneficial for improving clinical decision making.