Kidney Cancer Staging: Deep Learning Neural Network Based Approach
Nathan Hadjiyski
Abstract
Kidney cancer is a common type of cancer that can be very deadly. Proper cancer staging is critical for the correct selection of treatment for the affected patients. Stage 1 kidney cancer is an important threshold for the treatment decision. The physicians usually have difficulty to determine the cancer stage correctly, which can result in under- or overtreatment. Deep Learning Neural Network (DLNN) was used to predict kidney cancer Stage 1 versus higher stages in order to allow for a more accurate kidney cancer stage assessment by physicians. Computer tomography (CT) scans from 227 patients with different stages kidney cancer from the Cancer Imaging Archive TCIA database were used for training, validation, and testing of the DLNN. The kidney cancers were cropped from the 3D CT scans. The dataset was split into 48% training, 10% validation, and 42% test sets. Inception V3 DLNN with transfer learning was trained with the cropped kidney cancer training images. Area under the ROC curve (AUC) was used to estimate the classification accuracy. The AUC of 0.97 for training, 0.91 for validation and 0.90 for the test sets was obtained. This AI system shows promise for potentially assisting physicians in kidney cancer staging.