Noninvasive Pathological Staging of Clear Cell Renal Cell Carcinoma using Computed Tomography-based Radiomics Features and Machine Learning

According to Global Burden of Cancer data 2018, kidney cancer is one of the cancer cases that causes 9.6 million deaths in 185 countries. Clear cell renal cell carcinoma (ccRCC) is the most common type of RCC cancer. Accurate staging of ccRCC is important for determining prognosis and formulating effective treatment. A radiomics approach is proposed as non-invasively I to IV pathological staging of ccRCC using computed tomography (CT) images. 190 patients with pathological ccRCC stages I to IV were included. The research process consists of: patient selection, cancer mask segmentation, feature extraction and selection, and classification. 33 radiomics features were selected and used for system training using support vector machine (SVM) and provided the accuracy, sensitivity, and specificity of the tested dataset are 90%, 90%, and 96.67%, respectively. These results indicate that the pathological stage of ccRCC can be predicted significantly based on radiomics features on CT images. Radiomics features on 3D CT images can incredibly predict four pathological stages of ccRCC non-invasively. And machine learning can also be used to analyze potential feature sets and construct classification systems. To increase system reliability, potential directions of secured data sharing technology is also discussed in this study.