Objective: To investigate the feasibility of contrast-enhanced computer tomography (CT) texture analysis in predicting early recurrence after transarterial chemoembolization (TACE) in patients with liver cancer. Methods: A retrospective analysis was performed for 47 patients with liver cancer confirmed by liver biopsy and digital subtraction angiography who underwent upper abdominal contrast-enhanced CT scan before TACE, and according to the presence or absence of focal recurrence within half a year, these patients were divided into early recurrence (ER) group and non-early recurrence (NER) group. The texture analysis was used to delineate tumor boundary layer by layer on the axial contrast-enhanced CT image before liver cancer surgery, and related parameters of tumor heterogeneity, including entropy, mean, non-uniformity, skewness, and kurtosis, were obtained. The independent samples t-test was used for comparison of texture parameters between the two groups. The receiver operating characteristic (ROC) curve was used for the analysis of entropy, mean, and non-uniformity, and the area under the ROC curve (ROC), optical cut-off value, sensitivity, and specificity were calculated to evaluate the efficiency of texture analysis in predicting early focal recurrence after TACE. Results: There were 20 patients in the ER group and 27 in the NER group. The ER group had a maximum major axis length of 88.2±36.3 mm and a maximum minor axis length of 41.4±21.4 mm, and the NER group had a maximum major axis length of 66.9±30.2 mm and a maximum minor axis length of 29.3±19.8 mm; the ER group had significantly higher maximum major and minor axis lengths than the NER group (t = 4.89 and 4.62, P < 0.001). The ER group had significantly higher entropy and non-uniformity values than the NER group, and there were no significant differences in skewness and kurtosis between the two groups. Entropy, non-uniformity, and mean had high efficiency in predicting early recurrence after TACE, and the optimal cut-off value of entropy was 4.135. Conclusion: Volumetric texture analysis of contrast-enhanced CT images before liver cancer surgery has a high value in predicting early recurrence after TACE.
目的： 探讨CT增强纹理分析预测肝癌经动脉化学栓塞治疗（TACE）术后患者早期复发的可行性。 方法： 回顾性分析TACE术前行上腹部CT增强检查、经穿刺病理、数字减影血管造影证实的肝癌患者47例，以半年内有无局灶性复发为界，分为早期复发（ER）与非早期复发（NER）两组，应用纹理分析在肝癌术前CT轴位增强图像上逐层勾画肿瘤边界，统计得到两组患者整个瘤体异质性的相关参数：熵值、平均值、不均匀度、偏度及峰度等。采用独立样本t检验对两组间纹理分析参数进行比较，对熵值、均数及不均匀度三组参数进行受试者工作特征曲线分析，分别计算曲线下面积、最佳截断点、敏感度及特异度，评价其对TACE术后早期局灶性复发的预测效能。 结果： ER组20例，NER组27例。ER组肿瘤最大径线明显大于NER组，最大长径分别为（88.2± 36.3）mm、（41.4±21.4）mm，短径分别为（66.9±30.2）mm、（29.3±19.8）mm，差异有统计学意义（t值分别为4.89、4.62，P<0.001）。ER组中熵值及不均匀度明显高于NER组，均值在ER组中偏低；而两组间偏度和峰度差异无统计学意义。熵值、不均匀度及均数对TACE术后早期复发具有较高的预测效能，熵值最佳截断点为4.135。 结论： 通过术前肝癌CT增强图像的容积纹理分析，对TACE术后早期复发具有较高的预测价值。.
Keywords: Carcinoma, hepatocellular; Texture analysis; Transarterial chemoembolization.