基于影像组学与深度学习的食管癌磁共振成像的T分期

T staging of esophageal cancerwith MRI based on radiomics and deep learning

文中利用影像组学和深度学习方法在磁共振图像上进行食管癌T分期诊断。数据为244例食管癌患者r-VIBE序列磁共振图像。分别用影像组学和深度学习方法进行了食管癌T分期的建模,并进一步将两种方法结合建模。使用接收者操作特性曲线(ROC)和曲线下面积(AUC)对模型性能进行评价。影像组学模型的测试集AUC=0.765,深度学习模型验证集AUC=0.777,测试集AUC=0.703,两种方法组合模型测试集AUC=0.783。结果表明影像组学和深度学习均能用于基于MRI图像的食管癌T分期,结合使用两种方法能够取得更好的诊断效果。

This paper proposes radiomics and deep learning for automatic classification of T staging of esophageal cancer from MRI images.244 patients with esophageal cancer are included in this study.Both radiomics and deep learning methods are used for modeling,which are then combined to build a new model.Finally,the models are evaluated with receiver operating characteristic curve(ROC)and area under the curve(AUC).The test AUC of the radiomics model is 0.765,and the validation and test AUC of the deep learning model are 0.777 and 0.703 respectively.The test AUC of the final combined model is 0.783.The results demonstrate that both the radiomics and deep learning models can be used in MRI T staging of esophageal cancer,and better results can be achieved when these two methods are combined.