基于定位CT瘤内瘤周影像组学联合临床特征对肝转移瘤放疗后近期敏感性的研究
An Exploratory Study on Short-term Radiosensitivity of Liver Metastases after Radiotherapy Based on Intratumoral and Peritumoral Radiomics from Planning CT Combined with Clinical Features
尚文颖 1龙蝶 2斯韬 3陈海辉4
作者信息
- 1. 广西壮族自治区柳州市，柳州市妇幼保健院院长办公室;广西壮族自治区柳州市，广西中医药大学第三附属医院（柳州市中医医院）肿瘤科
- 2. 广西壮族自治区柳州市，广西医科大学第四附属医院（柳州市工人医院）肿瘤内科
- 3. 广西壮族自治区柳州市，广西中医药大学第三附属医院（柳州市中医医院）肿瘤科
- 4. 广西壮族自治区柳州市，柳州市妇幼保健院院长办公室
- 折叠
摘要
背景 转移性肝癌（MLC）多由消化道肿瘤等转移所致，立体定向放疗（SBRT）是其重要局部治疗手段，但肝脏肿瘤异质性导致患者放疗反应差异显著，筛选放疗获益优势人群成为临床关键难题。影像组学可从医学影像中提取大量可量化特征，在肿瘤诊疗中具有显著优势，但现有研究多基于诊断性CT或MRI，与放疗定位图像存在采集参数、时相差异。且研究多基于瘤内特征研究，缺少瘤周特征的系统性对比。目的 探索基于定位CT瘤内与瘤周影像组学特征联合临床因素预测肝转移瘤放疗近期疗效的价值。方法 回顾性收集2018年1月—2025年4月柳州市中医医院院接受放疗的120例肝转移瘤患者，按7:3随机分为训练集（n=84）例与测试集（n=36）例。采用多因素Logistic回归分析筛选临床独立影响因素；提取瘤内、瘤周3mm及5 mm区域影像组学特征，采用最小绝对收缩和选择算子（LASSO）筛选特征，并使用随机森林（RF）算法构建肝转移瘤放疗近期疗效预测模型。以放疗结束后3个月的实体瘤疗效评价标准（RECIST 1.1）版评估近期疗效，将完全缓解和部分缓解纳入放疗敏感，疾病稳定和疾病进展纳入放疗抵抗。通过受试者工作特征（ROC）曲线、DeLong 检验评估模型的预测效能及效能差异。结果 多因素Logistic回归分析显示，肝转移瘤数量是短期放疗疗效的独立影响因素（OR=0.016，95%CI=0.003~0.083，P<0.001）。LASSO回归筛选出与疗效密切相关的5个影像组学特征用于预测模型构建，其中 INTRA、Peri3mm、Peri5mm、IntraPeri5mm、IntraPeri3mm、Clinic预测模型预测肝转移瘤放疗近期疗效的AUC分别为0.767、0.700、0.689、0.803、0.828、0.783。IntraPeri3mm融合Clinic的联合模型（Combined）AUC为0.864，准确度为0.889，灵敏度0.833，特异度0.900。DeLong检验结果显示，IntraPeri3mm模型优于单一模型（P<0.05）。结论 CT瘤内联合瘤周3 mm影像组学特征融合临床因素可有效预测肝转移瘤放疗近期疗效，具有良好的临床转化潜力。
Abstract
Background Metastatic liver cancer (MLC) is most commonly caused by metastasis from gastrointestinal tumors. Stereotactic body radiotherapy (SBRT) is an important local treatment modality; however, tumor heterogeneity in the liver leads to substantial variability in patient responses, making the identification of patients most likely to benefit from radiotherapy a critical clinical challenge. Radiomics enables the extraction of numerous quantifiable features from medical images and offers significant advantages in tumor diagnosis and treatment. However, most existing studies are based on diagnostic CT or MRI, which differ in acquisition parameters and timing from radiotherapy planning images. Moreover, most studies focus on intratumoral features, with a lack of systematic comparison involving peritumoral features. Objective To explore the value of combining intratumoral and peritumoral radiomic features derived from planning CT with clinical factors in predicting the short-term efficacy of radiotherapy for liver metastases. Methods A retrospective analysis was conducted on 120 patients with liver metastases who received radiotherapy at Liuzhou Traditional Chinese Medicine Hospital from January 2018 to April 2025. Patients were randomly divided into a training cohort (n=84) and a test cohort (n=36) at a 73 ratio. Multivariate Logistic regression analysis was used to identify independent clinical factors. Radiomic features were extracted from the intratumoral region and peritumoral regions with 3 mm and 5 mm margins. Features were selected using the least absolute shrinkage and selection operator (LASSO) regression, and prediction models were constructed using the random forest (RF) algorithm. Short-term efficacy was evaluated using the Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1 at three months post-radiotherapy; patients with complete response and partial response were classified as radiosensitive, while those with stable disease and progressive disease were classified as radioresistant. Model performance was assessed using receiver operating characteristic (ROC) curves and the DeLong test. Results Multivariate Logistic regression analysis identified the number of liver metastases as an independent factor affecting short-term radiotherapy response (OR=0.016, 95%CI=0.003-0.083, P<0.001). LASSO regression selected five radiomic features strongly associated with efficacy for predictive model construction. The AUCs for predicting short-term radiotherapy efficacy of liver metastases were 0.767, 0.700, 0.689, 0.803, 0.828, and 0.783 for the INTRA, Peri3mm, Peri5mm, IntraPeri5mm, IntraPeri3mm, and Clinic models, respectively. The combined model (IntraPeri3mm-Clinic) achieved an AUC of 0.864, with an accuracy of 0.889, a sensitivity of 0.833, and a specificity of 0.900. Delong's test indicated that the IntraPeri3mm model was superior to the single models (P<0.05). Conclusion CT-based radiomic features integrating intratumoral and peritumoral (3 mm) regions, combined with clinical factors, can effectively predict the short-term response of liver metastases to radiotherapy, demonstrating promising potential for clinical translation.关键词
肝转移/放射治疗/影像组学/瘤周区域/预测模型Key words
Liver metastases/Radiotherapy/Radiomics/Peritumoral region/Prediction model引用本文复制引用
尚文颖,龙蝶,斯韬,陈海辉.基于定位CT瘤内瘤周影像组学联合临床特征对肝转移瘤放疗后近期敏感性的研究[EB/OL].(2026-06-29)[2026-07-03].https://chinaxiv.org/abs/202607.00004.学科分类
肿瘤学/医学研究方法
