Zhou Yuntao, Xiao Li, Yang Siyi, Yang Chengwen, Sun Jifeng, Wu Jiehan, Cui Zhiyong, Zhao Lujun, Sun Yunchuan, Liu Ningbo
Department of Radiotherapy, Tianjin Medical University Cancer Institute & Hospital, National Clinical Research Center for Cancer, Tianjin's Clinical Research Center for Cancer, Key Laboratory of Cancer Prevention and Therapy, Tianjin, Tianjin, China.
Department of Radiation Oncology, Hebei Province Cangzhou Hospital of Integrated Traditional and Western Medicine, Cangzhou, China.
Transl Lung Cancer Res. 2025 Jul 31;14(7):2584-2597. doi: 10.21037/tlcr-2025-326. Epub 2025 Jul 28.
Limited-stage small-cell lung cancer (LS-SCLC) is highly aggressive and prone to brain metastasis (BM). Early identification of BM risk is crucial for devising personalized prophylactic cranial irradiation (PCI) strategies. This study aimed to develop a multimodal model integrating radiomic and clinical features to stratify BM risk in LS-SCLC patients and guide personalized PCI strategies.
This study analyzed 141 LS-SCLC patients (2013-2021) using computed tomography (CT) images and clinical records. Patients were randomly divided into training (n=98), internal validation (n=43), and external validation cohorts (n=24). Radiomic features were extracted and optimized using the minimum redundancy maximum relevance (mRMR) algorithm to form a radiomic score (RadScore). Clinical predictors were identified via univariate logistic regression (LR). Four machine learning models-LR, support vector machine, random forest, and eXtreme Gradient Boosting-were used to develop predictive models. Model performance was evaluated by the area under the receiver operating characteristic curve (AUC).
A total of 141 patients (mean age, 59.03 years; 109 men and 32 women) were evaluated. A total of 1,037 radiomic features were extracted from the simulated positioning CT images, with 10 optimal features selected to form the RadScore. By incorporating dynamic changes in platelet count, hemoglobin levels, and leukocyte indices before and after radiotherapy, along with the baseline lymphocyte-to-monocyte ratio (LMR), the LR combined model demonstrated superior predictive capability. The LR combined model showed superior performance with AUCs of 0.831 (training), 0.831 (internal validation), and 0.863 (external validation). Risk stratification indicated that PCI reduced BM risk in high-risk patients [hazard ratio (HR) =0.270, P<0.001] but not in low-risk patients (HR =0.225, P=0.13).
The LR combined radiomic-clinical model demonstrated superior predictive performance. PCI significantly reduced the risk of BM in high-risk patients, whereas no statistically significant benefit was observed in low-risk patients.
局限期小细胞肺癌（LS-SCLC）侵袭性强，易发生脑转移（BM）。早期识别BM风险对于制定个性化的预防性颅脑照射（PCI）策略至关重要。本研究旨在开发一种整合放射组学和临床特征的多模态模型，以对LS-SCLC患者的BM风险进行分层，并指导个性化PCI策略。
本研究分析了141例LS-SCLC患者（2013年至2021年）的计算机断层扫描（CT）图像和临床记录。患者被随机分为训练组（n = 98）、内部验证组（n = 43）和外部验证组（n = 24）。使用最小冗余最大相关（mRMR）算法提取并优化放射组学特征，以形成放射组学评分（RadScore）。通过单因素逻辑回归（LR）确定临床预测因素。使用四种机器学习模型——逻辑回归、支持向量机、随机森林和极端梯度提升——来开发预测模型。通过受试者操作特征曲线（AUC）下的面积评估模型性能。
共评估了141例患者（平均年龄59.03岁；男性109例，女性32例）。从模拟定位CT图像中提取了总共1037个放射组学特征，选择10个最佳特征形成RadScore。通过纳入放疗前后血小板计数、血红蛋白水平和白细胞指数的动态变化，以及基线淋巴细胞与单核细胞比值（LMR），逻辑回归联合模型显示出卓越的预测能力。逻辑回归联合模型表现出色，训练组、内部验证组和外部验证组的AUC分别为0.831、0.831和0.863。风险分层表明，PCI降低了高危患者的BM风险[风险比（HR）= 0.270，P < 0.001]，但未降低低危患者的BM风险（HR = 0.225，P = 0.13）。
逻辑回归联合放射组学-临床模型显示出卓越的预测性能。PCI显著降低了高危患者的BM风险，而在低危患者中未观察到统计学上的显著获益。
