Dong Sixue, Yao Zian, Zhang Zhiyuan, Wang Jiazhou, Guo Ying, Hu Weigang, Ou Xiaomin, Hu Chaosu
Department of Radiation Oncology, Fudan University Shanghai Cancer Center, Shanghai 200032, China; Department of Oncology, Shanghai Medical College, Fudan University, Shanghai 200032, China; Shanghai Clinical Research Center for Radiation Oncology, Shanghai 200032, China; Shanghai Key Laboratory of Radiation Oncology, Shanghai 200032, China.
Department of Radiation Oncology, Fudan University Shanghai Cancer Center, Shanghai 200032, China; Department of Oncology, Shanghai Medical College, Fudan University, Shanghai 200032, China; Shanghai Clinical Research Center for Radiation Oncology, Shanghai 200032, China; Shanghai Key Laboratory of Radiation Oncology, Shanghai 200032, China.
Appl Radiat Isot. 2025 Aug 13;226:112100. doi: 10.1016/j.apradiso.2025.112100.
This study aims to propose a method for building a prognostic model with small sample sizes and to develop a predictive model combining radiomics and dosiomics for patients with locally advanced hypopharyngeal cancer treated with postoperative chemoradiotherapy, to improve prognostic accuracy despite the challenge of limited data.
A retrospective cohort of 48 male patients diagnosed with locally advanced hypopharyngeal cancer was included in the study. Radiomics features were extracted from pre-surgical MRI scans, and dosiomics features were derived from dose-volume histograms. To mitigate the small sample size issue, synthetic data generation techniques were employed to expand the dataset. A support vector machine (SVM) classifier was used to develop predictive models for progression-free survival (PFS) and overall survival (OS). Kaplan-Meier survival curves were used to stratify patients into high-risk and low-risk groups based on the risk scores. Additionally, univariate and multivariate Cox proportional hazards regression analyses were performed to evaluate the relationship between the risk scores and survival outcomes.
The combined radiomics and dosiomics model outperformed the radiomics-only model in predicting both PFS and OS. The integration of synthetic data significantly improved the model's robustness and generalizability. For PFS prediction, the model achieved AUCs of 0.92, 0.90, and 0.83 at 24, 30, and 36 months in the validation cohort, respectively. The model showed promising results for OS prediction as well, with improved accuracy when synthetic data was included. The performance of the Cox model and the accuracy of survival stratification was significantly improved with the omics-based risk score. The univariate analysis confirmed the independent prognostic value of the risk scores.
Synthetic data technique was confirmed to be effective for developing prognostic models in a small-sample cohort, proved that the combination of radiomics and dosiomics improved survival predictions for hypopharyngeal cancer patients treated with postoperative chemoradiotherapy.
本研究旨在提出一种构建小样本预后模型的方法，并为接受术后放化疗的局部晚期下咽癌患者开发一种结合放射组学和剂量组学的预测模型，以在数据有限的挑战下提高预后准确性。
本研究纳入了48例诊断为局部晚期下咽癌的男性患者的回顾性队列。从术前MRI扫描中提取放射组学特征，并从剂量体积直方图中导出剂量组学特征。为缓解小样本量问题，采用合成数据生成技术来扩充数据集。使用支持向量机（SVM）分类器开发无进展生存期（PFS）和总生存期（OS）的预测模型。采用Kaplan-Meier生存曲线根据风险评分将患者分为高风险和低风险组。此外，进行单变量和多变量Cox比例风险回归分析以评估风险评分与生存结果之间的关系。
在预测PFS和OS方面，联合放射组学和剂量组学模型优于仅放射组学模型。合成数据的整合显著提高了模型的稳健性和通用性。对于PFS预测，该模型在验证队列中24、30和36个月时的AUC分别为0.92、0.90和0.83。该模型在OS预测方面也显示出有前景的结果，纳入合成数据时准确性有所提高。基于组学的风险评分显著提高了Cox模型的性能和生存分层的准确性。单变量分析证实了风险评分的独立预后价值。
合成数据技术被证实对在小样本队列中开发预后模型有效，证明放射组学和剂量组学的结合改善了接受术后放化疗的下咽癌患者的生存预测。
