摘要： 目的 探讨基于磁共振T2加权脂肪抑制成像（T2WI-FS）的影像组学对前列腺癌盆腔淋巴结转移的诊断价值。方法 收集2014年12月至2019年12月中国医科大学附属第一医院经病理证实的前列腺癌患者42例（训练组29例，验证组13例）的临床资料，患者术前均行前列腺磁共振成像（MRI）平扫、动态对比增强MRI（DCE-MRI）及弥散加权成像（DWI）检查。在T2WI-FS影像上选取盆腔淋巴结感兴趣区，采用LASSO回归降维筛选出对前列腺癌转移与未转移淋巴结最具鉴别意义的纹理特征参数后构建影像组学模型，分别于训练组和验证组中对模型的预测效能进行分析。结果 采用LASSO回归在前列腺癌患者T2WI-FS影像中筛选出9个盆腔淋巴结转移最具鉴别意义的纹理特征参数（均P<0.01），分别为S（3，-3）SumEntrp、WavEnLH_s-4、Vertl_RLNonUni、Entropy、Horzl_GLevNonU、135dr_Fraction、45dr_RLNonUni、Teta3和S（0，1）SumOfSqs。通过logistics回归模型构建基于T2WI-FS的影像组学模型，训练组的受试者工作特征（ROC）曲线下面积（AUC）为0.922（95% CI：0.850~0.994，P<0.05），灵敏度和特异度分别为86.9%、84.3%；验证组的AUC为0.885（95% CI：0.804~0.966，P<0.05），灵敏度和特异度分别为81.8%、85.1%。结论 基于磁共振T2WI-FS的影像组学模型对前列腺癌盆腔淋巴结转移具有较高的诊断效能，能够为前列腺癌盆腔淋巴结切除术提供影像学依据。

基于磁共振T2WI-FS的影像组学对前列腺癌盆腔淋巴结转移的诊断价值

赖树盛¹, 郑石磊²

1. 中国医科大学附属第一医院放射科, 沈阳 110001;
2. 锦州医科大学附属第一医院放射科, 辽宁 锦州 121001

收稿日期:2020-08-31出版日期:2021-03-30发布日期:2021-03-20
通讯作者:郑石磊E-mail:114305970@qq.com
作者简介:赖树盛(1984-),男,技师,大专.
基金资助:辽宁省科技厅博士科研启动基金（2019-BS-099）


关键词: 磁共振成像, 影像组学, 前列腺癌, 淋巴结转移
Abstract: Objective To analyze the diagnostic value of magnetic resonance imaging (MRI) radiomic features of fat-suppressed T2-weighted imaging (T2WI-FS) for pelvic metastatic and non-metastatic lymph nodes in prostate cancer. Methods The data of 42 patients with pathologically confirmed prostate cancer were retrospectively collected from December 2014 to December 2019 (29 and 13 patients were assigned to the training and validation groups,respectively). All the patients underwent MRI examination including plain scan,dynamic contrast-enhanced MRI (DCE-MRI),and diffusion-weighted imaging (DWI) before surgery. Regions of interest (ROI) in the pelvic lymph nodes were selected on the T2WI-FS images,and the texture features that were significant for differentiating the metastatic from the non-metastatic lymph nodes in prostate cancer were selected by LASSO regression. Receiver operating characteristic (ROC) curves were used to assess the diagnostic performance of the radiomics models on the training and validation datasets. Results LASSO regression was used to extract nine texture features that were most significant for differentiating metastatic from non-metastatic pelvic lymph nodes in prostate cancer from T2WI-FS images. The texture parameters include S (3,-3) SumEntrp,WavEnLH_s-4,Vertl_RLNonUni,Entropy,Horzl_GLevNonU,135dr_Fraction,45dr_RLNonUni,Teta3,and S (0,1) SumOfSqs (all P<0.01). The area under the curve (AUC) for the training group was 0.922 (95% CI:0.850-0.994,P<0.05). The sensitivity and specificity were 86.9% and 84.3%,respectively. For the validation group,the AUC was 0.885 (95% CI:0.804-0.966,P<0.05),and the sensitivity and specificity were 81.8% and 85.1%,respectively. Conclusion The radiomics model based on T2WI-FS showed high diagnostic performance for pelvic metastatic lymph nodes in prostate cancer,and it can be used to provide imaging evidence for pelvic lymphadenectomy in prostate cancer.
Key words: magnetic resonance imaging, radiomics, prostatic cancer, lymphatic metastasis
PDF全文下载地址:

https://journal.cmu.edu.cn/CN/article/downloadArticleFile.do?attachType=PDF&id=2713