Objective: To explore the feasibility of short-term efficacy prognosis prediction model for HCC patients undergoing transcatheter arterial chemoembolization (TACE) based on MRI-based radiomics technique. Methods: A total of 123 patients with liver cancer who received TACE treatment in Lishui Central Hospital from June 2016 to July 2018 were retrospectively collected, including 90 males and 33 females, with an average age of 24-83 (58±10) years. All the patients were pathologically confirmed as hepatocellular carcinoma and underwent MRI scan before surgery.All patients were followed up 3-4 months after TACE, and further divided into training group (n=85, 42 of which were effective and 43 cases were ineffective) and the validation group (n=38, 19 of which were effective and 19 were ineffective) according to the modified response evaluation criteria in solid tumors (mRECIST). There was no statistical difference in the general information between the two groups of patients, which was comparable. Then, preoperative T(2)WI images were used for radiomics analysis, texture parameters were screened based on R language, and short-term efficacy prediction model of TACE for training group and verification group was constructed. Results: T(2)WI image analysis of each patient received 396 different texture parameters, and further used Lasso dimensionality reduction and 10 times cross-validation screening to obtain 5 characteristic texture parameters, specifically stdDeviation, ClusterProminence_angle135_offset4, Correlation_angle135_offset4, Inertia_angle135_offset4, InverseDifferenceMoment_angle45_offset4. According to the above five texture parameters and their corresponding coefficient values, the corresponding radiomics scores (Radscore) were calculated, and the prediction models of the training group and the verification group were further constructed.It was found that the area under the ROC curve of the training group was 0.812 (95%CI: 0.722-0.901), the sensitivity and specificity were 83.7% and 69.0%, respectively. The area under the ROC curve of the validation group was 0.801 (95%CI:0.654-0.947), and the sensitivity and specificity were 89.5% and 63.2%, respectively. Conclusion: The constructed TACE prediction model in the present study has high prediction accuracy, sensitivity and specificity.The short-term efficacy prognosis prediction model for HCC based on MRI is constructed, stable and reliable.
目的： 探索基于磁共振（MRI）的影像组学技术构建行经肝动脉化疗栓塞术（TACE）的肝癌患者短期疗效预后预测模型的可行性。 方法： 回顾性分析2016年6月至2018年7月在丽水市中心医院行TACE治疗的肝癌患者123例，其中男90例、女33例，年龄24~83岁，平均（58±10）岁，所有患者均经病理证实为肝细胞癌，并在术前行MRI扫描。同时，所有患者在TACE术后3~4个月进行影像学随访，并根据修正后实体瘤疗效评价标准（mRECIST）进一步分为训练组（85例，其中42例有效，43例无效）和验证组（38例，其中19例有效，19例无效），两组患者组内一般资料比较差异均无统计学意义，具有可比性。随后，利用术前T(2)WI图像进行影像组学分析，基于R语言筛选特征性纹理参数，构建训练组和验证组的TACE短期疗效预测模型。 结果： 每个患者的T(2)WI图像分析共得到396个不同的纹理参数，进一步利用LASSO降维及10倍交叉验证筛选得到5个特征性纹理参数，具体为stdDeviation，ClusterProminence_angle135_offset4，Correlation_angle135_offset4，Inertia_angle135_offset4，InverseDifferenceMoment_angle45_offset4，根据以上5个纹理参数及其对应系数值计算得到相应放射值（Radscores），并进一步构建训练组和验证组的预测模型，发现其中训练组模型的受试者工作特征（ROC）曲线下面积为0.812（95%CI：0.722~0.901），敏感度和特异度分别为83.7%和69.0%，验证组模型的ROC曲线下面积为0.801（95%CI：0.654~0.947），敏感度和特异度分别为89.5%和63.2%。 结论： 本研究构建的TACE疗效预测模型具有较高的预测准确度、敏感度和特异度。利用基于MRI的影像组学技术预测肝癌TACE短期疗效是可行的，预测模式稳定且可靠。.
Keywords: Embolization,therapeutic; Liver neoplasms; Magnetic resonance imaging; Texture analysis.