This paper presents a convolutional neural network (CNN) approach for segmenting gigapixel pathology images into normal and cancerous pixels to aid breast cancer diagnosis. Each year, the treatment decisions for more than 230, 000 patients in the U.S. hinges on whether the cancer has metastasized away from the breast. Metastasis detection is currently performed by pathologists reviewing large expanses of biological tissues. This is labor intensive and error-prone. We present an automated approach to detect and localize tumors as small as 100×100 pixels in digitized microscopy images sized 100, 000×100, 000 pixels or larger. Our method leverages the Inception (V3) neural network architecture and obtains state-of-the-art results on the Camelyon16 dataset in the challenging lesion-level tumor detection task. At 8 false positives per image, we detect 92.4% of the tumors, relative to 82.7% by the previous best automated approach. For comparison, a human pathologist achieved 73.2% sensitivity using exhaustive search. We also achieved an image-level AUC above 97% on both the Camelyon16 test set, and another independent set of 110 slides. In addition, we discovered two Camelyon16 slides in the training set that were erroneously labeled normal. Our approach could considerably reduce false negative rates in metastasis detection.