Unsupervised Anomaly Detection on Histopathology Images Using Adversarial Learning and Simulated Anomaly

Automated analytics in computational histopathology have shown significant progress in aiding pathologists through digital image analysis. However, developing a robust model based on supervised learning for histopathology images is challenging because of the scarcity of tumor-marked samples and unknown diseases. Unsupervised anomaly detection (UAD) methods that were mostly used in industrial inspection are thus proposed to facilitate efficient analytics. UAD only requires normal samples for training and largely reduces the burden of labeling. In this paper, we introduce a reconstruction-based UAD approach called PathUAD to improve representation learning based on adversarial learning and simulated anomalies. On the one hand, we mix up features extracted from normal images to build a smoother feature distribution and employ adversarial learning to enhance an autoencoder for image reconstruction. On the other hand, we simulate anomalous images by image deformation, and guide the autoencoder to catch global characteristics of normal images well. We demonstrate its effectiveness on a histopathology anomaly detection benchmark and show state-of-the-art performance.