A radiomics-based machine learning pipeline to distinguish between metastatic and healthy bone using lesion-center-based geometric regions of interest; dataset

Please read the README file for more details.<br>`featurespace_metadata.json` includes radiomic features extracted from 1273 spinal lesions (healthy or metastatic) from radiotherapy planning-ct images using single point-based geometrical regions of interest (ROIs).<br>`output` is a folder containing the results of our radiomic-based machine learning pipeline in differentiating between healthy bone (HB) and bone metastases (BM) lesions. The pipeline was trained and tested using several resampling techniques (RS), feature selection methods (FS), and machine learning classifiers (ML) on single-point-based geometric ROIs with various shapes and sizes.

如需了解更多细节，请阅读README文件。`featurespace_metadata.json` 包含了从1273例脊柱病变（正常或转移性病变）的放疗计划CT图像中，基于单点几何感兴趣区域（Regions of Interest, ROIs）提取的放射组学特征。`output` 文件夹存储了我们基于放射组学的机器学习流程在区分正常骨（Healthy Bone, HB）与骨转移性病变（Bone Metastases, BM）时的实验结果。该机器学习流程依托不同形状与尺寸的单点几何ROIs，结合多种重采样技术（Resampling Techniques, RS）、特征选择方法（Feature Selection Methods, FS）以及机器学习分类器（Machine Learning Classifiers, ML）完成了模型的训练与测试。