基于术前CT图像的放射组学模型预测脊柱结核手术治疗预后的研究  

Study of preoperative CT-based radiomics models for the prediction of prognosis of surgical treatment of spinal tuberculosis

作　　者：宋敏[1] 谢智恩[1] 杨宏志[1] 吴华强 官权 钟鹏 方伟军[1] Song Min;Xie Zhien;Yang Hongzhi;Wu Huaqiang;Guan Quan;Zhong Peng;Fang Weijun(Department of Radiology,Guangzhou Chest Hospital,Guangzhou 510095,China)

机构地区：[1]广州市胸科医院放射科,广州510095

出　　处：《新发传染病电子杂志》2022年第2期52-56,共5页Electronic Journal of Emerging Infectious Diseases

基　　金：广东省医学科学技术研究基金项目(B2018224)。

摘　　要：目的构建基于脊柱结核患者术前CT特征的放射组学模型,预测脊柱结核手术治疗的预后,以提高术前诊断和个性化治疗的准确性。方法回顾性分析2015年1月至2019年5月在广州市胸科医院接受初次手术治疗的216例脊柱结核患者的临床及影像资料,按脊柱结核手术治疗1年的疗效,分为预后差组39例和临床治愈组177例。按7∶3的比例分别从预后差组及临床治愈组中随机抽取患者进入训练组及验证组(训练组:预后差组27例,临床治愈组124例;验证组:预后差组12例,临床治愈组53例)。将所有病例的术前平扫CT图像转换并勾画出两种模式的ROI,分别是病变椎体ROI、病变椎体及椎旁脓肿ROI,保存为组学处理文件并提取放射组学特征。利用L1正则化Logistic回归模型确定用于构建放射模型的最佳放射组学特征。利用ROC曲线的曲线下面积(AUC)评价所构建的放射组学模型的性能。结果椎体模型在训练队列中的AUC为0.892(95%CI:0.826~0.914),在验证队列中的AUC为0.851(95%CI:0.761~0.887)。椎体及椎旁脓肿模型在训练队列中的AUC为0.906(95%CI:0.809~0.920),在验证队列中的AUC为0.868(95%CI:0.786~0.904)。两个模型的预测效能比较差异均无统计学意义(均P>0.05)。结论基于脊柱结核术前CT图像的放射组学模型,能够有效预测患者脊柱结核手术治疗预后,有助于脊柱结核临床治疗方法的选择。Objective Construct radiomics models based on preoperative CT features of patients with spinal tuberculosis to predict the prognosis of surgical treatment,so as to improve the accuracy of preoperative diagnosis and personalized treatment.Method The clinical and imaging data of 216 patients with spinal tuberculosis who received initial surgical treatment in Guangzhou Chest Hospital from January 2015 to May 2019 were analyzed retrospectively.According to the prognosis one year after operation,they were divided into Bad Group(39 cases)and Good Group(177 cases).The patients in Bad Group and Good Group were randomly divided into training cohort and validation cohort with a ratio of 7:3(training cohort:27 cases in Bad Group and 124 cases in Good Group;validation cohort:12 cases in Bad Group and 53 cases in Good Group).The preoperative plain CT images of all the patients were converted and outlined two kinds of region of interest(ROI),including the ROI of diseased vertebra and the ROI of diseased vertebra with peripheral abscess,which were saved as radiomics files and extracted radiomics features.The L1 regularized logistic regression model was used to identify the optimal radiomics features for construction of radiomics models.The area under the curve(AUC)of the receiver operating characteristic curve(ROC)was used to evaluate the performance of the constructed radiomics model.Result In the training cohort,the AUC of vertebra model was 0.892(95%CI:0.826,0.914),the AUC of vertebra model in the validation cohort was 0.851(95%CI:0.761,0.887).In the training cohort,the AUC of diseased vertebra with peripheral abscess model was 0.906(95%CI:0.809,0.920).In validation cohort,AUC of diseased vertebra with peripheral abscess model was 0.868(95%CI:0.786,0.904).There was no significant difference in the performance between the two models(P>0.05).Conclusion The radiomics models based on preoperative CT images of spinal tuberculosis can effectively predict the postoperative prognosis,and are helpful for surgeons to choose the clinic

关 键 词：脊柱结核 手术治疗 放射组学 模型 预后 

分 类 号：R687.3[医药卫生—骨科学]