基于YAP流水线的影像组学研究流程的实现

Implement of radiomics flow based on the YAP pipeline

目的在又一条流水线(yet another pipeline,YAP)磁共振软件流水线框架基础上,增加对影像组学研究的支持。材料与方法在YAP中实现了对Python语言的支持,使得YAP流水线中可以嵌入Python语言编写的处理器;在此基础上,利用Py Radiomics软件包在YAP中实现了影像组学处理流水线。利用流水线,对BRATS2017公开数据集的高级别胶质瘤和低级别胶质瘤分级问题进行了研究。结果利用C++和Python混合编程,构建了完整的影像组学流水线,BRATS2017数据集的肿瘤分级在选择12个特征时,获得的最佳准确率达到了94.5%,受试者操作特性(receiver operating characteristic,ROC)曲线的曲线下面积(area under curve,AUC)达到了0.9650。结论利用Python和C++的混合编程以及YAP框架提供的基础设施,可以方便地进行影像组学研究。

Objective:To support radiomics studies based YAP(Yet Another Pipeline),which is originally a framework for magnetic resonance image reconstruction and post-processing.Materials and Methods:We introduced support of Python into YAP pipeline,so that processors in the pipeline can be programmed in Python.Then we implemented a radiomics pipeline with PyRadiomics package.Finally,the pipeline was used to study brain tumor grading problem with BRATS2017 open datasets.Results:A complete radiomics pipeline was built,which involved hybrid programming of C++and Python.Best results for BRATS2017 tumor grading were achieved when 12 features were selected,with the best accuracy of 94.5%and AUC(Area Under Curve)for receiver operating characteristic curve of 0.9650.Conclusions:Hybrid programming of Python and C++,together with the facilities provided by YAP framework,may facilitate radiomics studies.