Hepatocellular carcinoma (HCC) patients treated with sorafenib received CT imaging and clinical data analysis to develop a radiomics signature (RadSig1) predicting overall survival (OS) based on baseline and week 10 features. The model combined eight clinical and radiomic variables, showing strong association with OS in validation (HR=2398, P<0.001). It outperformed traditionalRECIST criteria and highlighted limitations of size-based metrics.分隔符
研究创新性地提出"双时点"影像组学分析框架:在基线及第10周两次CT扫描中,分别提取影像特征并计算其动态变化值。这种设计既避免了传统影像评估中"时间盲区"导致的 immortal time bias,又通过引入时间序列数据增强模型对治疗反应的敏感性。模型验证阶段发现,RadSig1可将患者分为四个风险等级,其中高危组(Q4)中位生存期仅为低危组(Q1)的7.3%,这种分层能力显著优于传统RECIST 1.1标准(总应答率仅7.7%)。