The variety of treatment options for clinically localized renal masses is diverse. Medical imaging depicts a non-invasive technique able to retrieve detailed information that can be utilized for treatment decision. For the ISBI KNIGHT challenge, we studied the ability of deep neural networks for renal mass risk score prediction based on CT imaging and clinical features. A U-Net archi-tecture was trained for segmentation of the region of interest. Afterwards, patches of both kidneys were input into the convolutional layers of a neural network, resulting feature vectors were fused using the element-wise maximum and clinical features were merged with deep features for risk score classification. The model achieved an area under the receiver operating characteristics curve (AUC) of 0.814 on the test set for dis-crimination of clinical relevant subclasses. For the more de-tailed risk score prediction task a mean AUC of 0.676 was achieved.