姬健智,张倩,牛猛,郭顺林,翟亚楠.联合临床、MR T2WI及表观弥散系数图影像组学特征列线图预测初发前列腺癌骨转移[J].中国医学影像技术,2022,38(7):1050~1055
联合临床、MR T2WI及表观弥散系数图影像组学特征列线图预测初发前列腺癌骨转移
Nomogram based on clinical, MR T2WI and apparent diffusion coefficient map radiomics features for predicting bone metastasis of incipient prostate cancer
投稿时间:2022-02-23  修订日期:2022-05-01
DOI:10.13929/j.issn.1003-3289.2022.07.020
中文关键词:  前列腺肿瘤  肿瘤转移  骨和骨组织  磁共振成像  影像组学  列线图
英文关键词:prostatic neoplasms  neoplasm metastasis  bone and bones  magnetic resonance imaging  radiomics  nomogram
基金项目:
作者单位E-mail
姬健智 兰州大学第一临床医学院, 甘肃 兰州 730000  
张倩 兰州大学第一临床医学院, 甘肃 兰州 730000  
牛猛 兰州大学第一医院放射科, 甘肃 兰州 730000  
郭顺林 兰州大学第一医院放射科, 甘肃 兰州 730000 guoshunlin@msn.com 
翟亚楠 兰州大学第一医院放射科, 甘肃 兰州 730000  
摘要点击次数: 149
全文下载次数: 74
中文摘要:
       目的 评估基于临床、MR T2WI及表观弥散系数(ADC)图影像组学特征构建的联合模型列线图预测初发前列腺癌骨转移的价值。方法 回顾性分析110例接受前列腺MR检查且经病理证实的初发前列腺癌患者,根据99Tcm亚甲基二磷酸盐(99Tcm-MDP)全身骨显像分为骨转移组(n=50)和无骨转移组(n=60)。基于T2WI及ADC图各提取851个、共1 702个影像组学特征,筛选最佳特征,计算影像组学评分并建立影像组学模型。应用单因素和多因素logistic回归分析筛选初发前列腺癌骨转移的临床相关独立危险因素,建立临床模型,并构建临床独立危险因素联合影像组学评分联合模型,绘制列线图将之可视化。以受试者工作特征(ROC)曲线评估各模型预测初发前列腺癌骨转移的效能,以决策曲线分析(DCA)评价联合模型的价值。结果 最终选出11个最佳影像组学特征,以之建立的影像组学模型预测初发前列腺癌骨转移的曲线下面积(AUC)为0.82。总前列腺特异性抗原、碱性磷酸酶和N分期是初发前列腺癌骨转移的临床独立危险因素(P均<0.05),以之构建的临床模型的AUC为0.93。联合模型的AUC(0.96)高于临床模型(Z=-2.066,P=0.039)和影像组学模型(Z=-3.451,P<0.001)。联合模型在阈值概率0~0.98时的临床净获益大于临床模型。结论 基于临床联合T2WI及ADC图影像组学特征的列线图可有效预测初发前列腺癌骨转移。
英文摘要:
      Objective To evaluate the value construct of a nomogram based on clinical, MR T2WI and apparent diffusion coefficient (ADC) map radiomics features for predicting bone metastasis of incipient prostate cancer. Methods Data of 110 patients with incipient prostate cancer confirmed by pathology who underwent prostate MR examinations were retrospectively analyzed. The patients were divided into bone metastasis group (n=50) and non-bone metastasis group (n=60) according to 99Tcm-methylene diphosphonate whole body skeletal imaging. Totally 1 702 radiomics features were extracted from T2WI (n=851) and ADC (n=851) maps, and the optimal features were selected to calculate the radiomics score and establish a radiomics model. Univariate and multivariate logistic regression analysis were used to screen the independent clinical risk factors of bone metastasis of incipient prostate cancer, and a clinical model was established. Then a combined model based on the independent clinical risk factors and radiomics score was constructed, and a nomogram was drawn to visualize the combined model. Receiver operating characteristic (ROC) curve was used to evaluate the efficacy of each model for predicting bone metastasis in incipient prostate cancer, and the value of the combined model was explored with decision curve analysis (DCA). Results Eleven optimal radiomics features were selected based on T2WI and ADC maps to establish a radiomics model, with the area under the curve (AUC) for predicting bone metastasis of incipient prostate cancer of 0.82. The total prostate specific antigen, alkaline phosphatase and N stage were all independent clinical risk factors of bone metastasis of incipient prostate cancer (all P<0.05), and AUC of the clinical model was 0.93. AUC of combined model (0.96) was higher than that of clinical model (Z=-2.066, P=0.039) and radiomics model (Z=-3.451, P<0.001). The combined model had greater clinical net benefit than the clinical model at the threshold probability of 0-0.98. Conclusion Nomogram based on clinical data, MR T2WI and ADC map radiomics features could effectively predict bone metastasis of incipient prostate cancer.
查看全文  查看/发表评论  下载PDF阅读器