肖奕,蒋昭琰,何舒欣,胡斯娴,文大光,张雨,李真林,夏春潮.T2WI、DWI及ADC图多序列影像组学模型评估直肠癌淋巴结转移[J].中国医学影像技术,2022,38(7):1029~1034
T2WI、DWI及ADC图多序列影像组学模型评估直肠癌淋巴结转移
T2WI, DWI and ADC map multi-sequence radiomics for evaluation on lymph node metastasis of rectal cancer
投稿时间:2021-09-29  修订日期:2022-04-22
DOI:10.13929/j.issn.1003-3289.2022.07.015
中文关键词:  直肠肿瘤  淋巴结转移  磁共振成像  影像组学
英文关键词:rectal neoplasms  lymphatic metastasis  magnetic resonance imaging  radiomics
基金项目:四川省科技计划项目(2019YFS0431)、四川大学华西医院学科卓越发展1·3·5工程项目(ZYGD18019)。
作者单位E-mail
肖奕 四川大学华西医院放射科, 四川 成都 610041  
蒋昭琰 四川大学华西医院放射科, 四川 成都 610041  
何舒欣 四川大学华西医院放射科, 四川 成都 610041  
胡斯娴 四川大学华西医院放射科, 四川 成都 610041  
文大光 四川大学华西医院放射科, 四川 成都 610041  
张雨 四川大学华西医院放射科, 四川 成都 610041  
李真林 四川大学华西医院放射科, 四川 成都 610041  
夏春潮 四川大学华西医院放射科, 四川 成都 610041 xiachunchao@126.com 
摘要点击次数: 206
全文下载次数: 104
中文摘要:
       目的 观察基于术前MR T2WI、弥散加权成像(DWI)及表观弥散系数(ADC)图多序列影像组学模型评估直肠癌淋巴结转移的价值。方法 回顾性分析74例经术后病理确诊单发直肠癌患者的T2WI、DWI和ADC图,按照7∶3比例将其分为训练集(n=52,21例淋巴结转移、31例无淋巴结转移)和测试集(n=22,9例淋巴结转移、13例无淋巴结转移)。由2名影像科医师以病理结果为标准基于常规MRI评价淋巴结转移,评估其诊断效能;分别基于T2WI、DWI、ADC图及三者联合提取病灶影像组学特征,筛选后构建影像组学模型,包括T2WI模型、DWI模型、ADC模型及多序列模型,并于训练集进行训练,于测试集评估其效能。绘制受试者工作特征(ROC)曲线,计算曲线下面积(AUC),观察各模型诊断直肠癌淋巴结转移的效能。结果 训练集与测试集患者性别、年龄、病灶位置、T分期及N分期差异均无统计学意义(P均>0.05)。常规MRI评估直肠癌淋巴结转移的准确率为58.11%(43/74),敏感度和特异度分别为76.67%(23/30)和45.45%(20/44)。T2WI、DWI、ADC及多序列模型评估测试集直肠癌淋巴结转移的AUC分别为0.78、0.68、0.77及 0.82,后者的准确率、敏感度及特异度分别为86.36%、88.89%及84.62%。结论 术前基于MR T2WI、DWI及ADC图多序列影像组学模型能有效评估直肠癌淋巴结转移。
英文摘要:
      Objective To observe the value of multi-sequence radiomics based on MR T2WI, diffusion weighted image (DWI) and apparent diffusion coefficient (ADC) map for evaluating lymph node metastasis of rectal cancer. Methods Preoperative data of T2WI, DWI and ADC maps of 74 patients with single rectal cancer confirmed by postoperative pathology were retrospectively analyzed. The patients were divided into training set (n=52, 21 with and 31 without lymph node metastasis) and testing set (n=22, 9 with and 13 without lymph node metastasis) at the ratio of 7 ∶ 3. Lymph node metastasis was evaluated by 2 radiologists based on conventional MRI, and its diagnostic efficacy was evaluated according to pathological results. Based on T2WI, DWI, ADC map and the combination of the three, the radiomics features of lesion were extracted and screened. Then radiomics models of T2WI, DWI, ADC model and the combination were constructed respectively and trained with training set, and the efficacies were evaluated with test set. Receiver operating characteristic (ROC) curves were drawn, and the area under the curve (AUC) were calculated to explore the efficacy of each model for predicting lymph node metastasis of rectal cancer. Results There was no significant difference of gender, age, lesion location, T stage nor N stage between training set and testing set (all P>0.05). The accuracy, sensitivity and specificity of lymph node metastasis of rectal cancer based on conventional MRI was 58.11% (43/74), 76.67% (23/30) and 45.45% (20/44), respectively. AUC of T2WI, DWI and ADC model for evaluating lymph node metastasis in rectal cancer was 0.78, 0.68 and 0.77, respectively, of multi-sequence model was 0.82, and the accuracy, sensitivity and specificity of the latter was 86.36%, 88.89% and 84.62%, respectively. Conclusion Multi-sequence radiomics model based on preoperative MR T2WI, DWI and ADC map could effectively evaluate lymph node metastasis of rectal cancer.
查看全文  查看/发表评论  下载PDF阅读器