宫颈癌放疗中基于精确表面剂量累加的直肠并发症预测模型
发布时间:2018-04-17 01:09
本文选题:宫颈癌放疗 + 直肠并发症 ; 参考:《南方医科大学学报》2017年12期
【摘要】:目的提出并验证一种应用于宫颈癌放疗的基于精确表面剂量累加的直肠并发症预测方法。方法回顾性采集到42例宫颈癌患者数据,对各个治疗分次的直肠壁进行精确点配准得到变形场,利用该变形场对直肠壁受照剂量进行变形和叠加,得到3D总剂量。将3D直肠总剂量分布映射到2D平面,分别提取出剂量体积特征和剂量几何特征,并筛选出有显著性差异(P0.05)的剂量分布特征,建立基于序列前向选择算法和逻辑回归的直肠并发症预测模型。结果直肠壁表面配准的精度较高,4个相似度评价指标表明,配准后不同治疗分次的直肠表面配准程度有显著性提高。五折交叉验证结果显示准确性、敏感性、特异性和AUC分别为:79.5%,81.3%,75.0%和0.88。结论本文提出基于精确表面剂量累加的直肠并发症预测模型具有可行性,为宫颈癌患者直肠毒性预测提供可靠的支持。
[Abstract]:Objective to propose and verify a method for predicting rectal complications based on accurate surface dose accumulation for cervical cancer radiotherapy.Methods data of 42 patients with cervical cancer were collected retrospectively. The deformation field was obtained by accurate registration of the rectum wall of each treatment grade. The total 3D dose was obtained by using the deformation field to deform and superposition the irradiated dose on the rectal wall.The 3D rectal total dose distribution was mapped to 2D plane to extract the dose volume characteristics and dose geometric characteristics, and to screen out the dose distribution characteristics with significant difference (P0.05).A prediction model of rectal complications based on sequential forward selection algorithm and logical regression was established.Results the accuracy of rectal surface registration was high. The four similarity evaluation indexes showed that the rectal surface registration degree of different treatment scores was significantly improved after registration.The accuracy, sensitivity, specificity and AUC were 75.0% and 0.88%, respectively.Conclusion the prediction model of rectal complications based on accurate surface dose accumulation is feasible and reliable for predicting rectal toxicity in patients with cervical cancer.
【作者单位】: 南方医科大学生物医学工程学院;
【基金】:国家自然科学基金(81728016,81571771) 广东省自然科学基金面上项目(2015A030313302) 广州市科技计划项目-珠江科技新星专项(201506010096)~~
【分类号】:R730.55;R737.33
【相似文献】
相关博士学位论文 前1条
1 王慧;胸外科患者术后慢性疼痛预测模型的建立[D];北京协和医学院;2014年
,本文编号:1761385
本文链接:https://www.wllwen.com/yixuelunwen/fangshe/1761385.html