Survival prediction in renal cell carcinoma patients using machine learning

Survival prediction and outcome analysis are essential for any cancer patient. It helps in patient management and treatment plans. Patients with early-stage diagnosis of cancer have a higher survival rate which reduces morbidity and mortality. In this paper, machine learning algorithms are used for survival prediction in clear cell Renal cell Carcinoma (ccRCC). Three slices from the primary tumor and 3 slices from metastasis are selected. The texture features are extracted from these selected slices using a gray level correlation matrix (GLCM), discrete cosine transform (DCT), and discrete wavelet transform (DWT). A patient’s survival is estimated by selecting the most relevant features that are highly correlating within a group. Contrast enhanced computed tomography (CECT) scans of 38 patients are used in this study. K-nearest neighborhood and decision tree classifiers outperform the state-of-the-art machine learning algorithms and provide ‘59.60%’ and ‘57.90%’ classification accuracy for survival prediction, respectively.