Detecting Cancer Metastases on Gigapixel Pathology Images

TLDR

Metastasis detection in breast cancer, which informs treatment decisions for over 230,000 patients annually, currently relies on labor‑intensive, error‑prone pathologist review of large tissue areas. The study proposes an automated framework to detect and localize tumors as small as 100 × 100 px in 100 000 × 100 000‑pixel gigapixel pathology images. The method uses a convolutional neural network architecture to achieve state‑of‑the‑art lesion‑level tumor detection on the Camelyon16 dataset. The approach detects 92.4 % of tumors at 8 false positives per image—outperforming the prior best 82.7 % and a human pathologist’s 73.2 % sensitivity—while achieving over 97 % image‑level AUC on Camelyon16 and an independent set, revealing two mislabeled slides and potentially lowering false‑negative rates.

Abstract

Each year, the treatment decisions for more than 230,000 breast cancer patients in the U.S. hinge on whether the cancer has metastasized away from the breast. Metastasis detection is currently performed by pathologists reviewing large expanses of biological tissues. This process is labor intensive and error-prone. We present a framework to automatically detect and localize tumors as small as 100 x 100 pixels in gigapixel microscopy images sized 100,000 x 100,000 pixels. Our method leverages a convolutional neural network (CNN) 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 attempting exhaustive search achieved 73.2% sensitivity. We achieve image-level AUC scores above 97% on both the Camelyon16 test set and an independent set of 110 slides. In addition, we discover that two slides in the Camelyon16 training set were erroneously labeled normal. Our approach could considerably reduce false negative rates in metastasis detection.