基于多模态特征和多分类器的宫颈癌放疗直肠毒性预测模型研究

放射治疗是宫颈癌的主要治疗方式之一，充分理解放射剂量分布与关键器官（如直肠）的放射毒性之间的关系，可以帮助放疗科医生有效、安全的提高靶区剂量，从而提高宫颈癌的局控率。但目前临床上沿用的剂量体积参数仍不能对关键器官的放射毒性进行有效的预测。申请人结合前期的研究工作基础，拟利用机器学习方法，构建一个融合多模态特征和多分类器的直肠放射毒性预测数学模型，以实现直肠放射毒性的个性化预测；并设计模型的反馈机制，分析与直肠放射毒性相关的解剖、临床、剂量学特征。研究内容：1）剂量形变累加的物理体模验证及算法优化；2）临床指标和剂量特征的分析、提取；3）融合多模态特征和多分类器的毒性预测模型的构建；4）模型验证和直肠毒性相关因素分析。本研究可以为宫颈癌放疗提供直肠毒性的个性化预测，并深入分析毒性相关的各种因素，辅助指导临床决策，具有较大的科学研究意义和临床应用价值。

Cervical cancer is usually treated with external beam radiotherapy followed by brachytherapy. Better understanding of the dose-toxicity relationship is critical for safe dose escalation to improve local control in cervical cancer radiotherapy. The dose-volume parameters, which are now commonly used in clinics, have been proved not to be the ideal predictors for organ at risk toxicity. Based on our previous works, this study is to propose an accurate rectal toxicity prediction model by fusing multimodality information and different classifiers. By designing novel model feedback mechanism, factors related to rectal toxicity, including clinical factors, anatomical factors and dosimetric factors, will be analyzed. This research includes: 1) accurate dose deformable accumulation and physical phantom evaluation; 2) clinical and dosimetric feature extraction and analysis; 3) multimodality features and classifiers fusion based rectal toxicity prediction modeling; and 4) model validation and rectal toxicity related factors analysis. This study will provide a personalized tool for rectal toxicity prediction in cervical cancer radiotherapy, and the in-depth investigations on factors relating to rectal toxicity will serve as guidance for clinical decision making in treatment planning for cervical cancer radiotherapy.