Wu Mingsong, Que Zenglong, Lai Shujie, Li Guanhui, Long Jie, He Yuqin, Wang Shunan, Wu Hao, You Nan, Lan Xiang, Wen Liangzhi
Department of Gastroenterology, Daping Hospital, Army Medical University (Third Military Medical University), No. 10, Changjiang Branch Road, Yuzhong District, Chongqing, 400042, P. R. China.
Department of Infectious Diseases, The 960 Hospital of PLA, No. 25, Shifan Road, Tianqiao District, Jinan City, Shandong Province, 250031, P. R. China.
Cell Oncol (Dordr). 2025 Feb 4. doi: 10.1007/s13402-025-01041-0.
Predicting the therapeutic response before initiation of hepatic artery infusion chemotherapy (HAIC) with fluorouracil, leucovorin, and oxaliplatin (FOLFOX) remains challenging for patients with unresectable hepatocellular carcinoma (HCC). Herein, we investigated the potential of a contrast-enhanced CT-based habitat radiomics model as a novel approach for predicting the early therapeutic response to HAIC-FOLFOX in patients with unresectable HCC.
A total of 148 patients with unresectable HCC who received HAIC-FOLFOX combined with targeted therapy or immunotherapy at three tertiary care medical centers were enrolled retrospectively. Tumor habitat features were extracted from subregion radiomics based on CECT at different phases using k-means clustering. Logistic regression was used to construct the model. This CECT-based habitat radiomics model was verified by bootstrapping and compared with a model based on clinical variables. Model performance was evaluated using the area under the curve (AUC) and a calibration curve.
Three intratumoral habitats with high, moderate, and low enhancement were identified to construct a habitat radiomics model for therapeutic response prediction. Patients with a greater proportion of high-enhancement intratumoral habitat showed better therapeutic responses. The AUC of the habitat radiomics model was 0.857 (95% CI: 0.798-0.916), and the bootstrap-corrected concordance index was 0.842 (95% CI: 0.785-0.907), resulting in a better predictive value than the clinical variable-based model, which had an AUC of 0.757 (95% CI: 0.679-0.834).
The CECT-based habitat radiomics model is an effective, visualized, and noninvasive tool for predicting the early therapeutic response of patients with unresectable HCC to HAIC-FOLFOX treatment and could guide clinical management and decision-making.
对于无法切除的肝细胞癌（HCC）患者，在开始使用氟尿嘧啶、亚叶酸钙和奥沙利铂（FOLFOX）进行肝动脉灌注化疗（HAIC）之前预测治疗反应仍然具有挑战性。在此，我们研究了基于对比增强CT的肿瘤栖息地放射组学模型作为预测无法切除的HCC患者对HAIC-FOLFOX早期治疗反应的新方法的潜力。
回顾性纳入了148例在三个三级医疗中心接受HAIC-FOLFOX联合靶向治疗或免疫治疗的无法切除的HCC患者。使用k均值聚类从不同阶段的CECT基于子区域放射组学中提取肿瘤栖息地特征。采用逻辑回归构建模型。这个基于CECT的肿瘤栖息地放射组学模型通过自举法进行验证，并与基于临床变量的模型进行比较。使用曲线下面积（AUC）和校准曲线评估模型性能。
识别出三种具有高、中、低强化的瘤内栖息地，以构建用于预测治疗反应的肿瘤栖息地放射组学模型。瘤内高强化栖息地比例更高的患者显示出更好的治疗反应。肿瘤栖息地放射组学模型的AUC为0.857（95%CI：0.798-0.916），自举校正一致性指数为0.842（95%CI：0.785-0.907），其预测价值优于基于临床变量的模型，后者的AUC为0.757（95%CI：0.679-0.834）。
基于CECT的肿瘤栖息地放射组学模型是预测无法切除的HCC患者对HAIC-FOLFOX治疗早期治疗反应的有效、可视化且非侵入性的工具，可指导临床管理和决策。
