ResNet-Based Model for Cancer Detection

Abstract

Cancer is a horrible disease and a major reason to cause death in the world. Early detection and diagnosis can help doctor save life. Many computer-aided diagnosis techniques use image processing to help doctor do cancer detection and obtain considerable achievements. In this paper, we propose a novel ResNet-based deep learning network to identify metastatic cancer from cancer scan images. Furthermore, we apply Test Time Augmentation to make our model more robust and improve detection accuracy. The results of experiments on a slightly modified version of the PatchCamelyon (PCam) benchmark dataset (the original PCam dataset contains duplicate images due to its probabilistic sampling, however, the version we use does not contain duplicates), which packs the clinically-relevant task of metastasis detection into a straightforward binary image classification task, indicates that the proposed ResNet-based model has achieved the state-of-the-art performance, which goes beyond performance of previous VGG16, VGG19 models.