SQUID: Deep Feature In-Painting for Unsupervised Anomaly Detection

TL;DR

SQUID is a novel deep learning method that uses space-aware memory inpainting to identify anomalies in radiography images by leveraging recurrent anatomical patterns, outperforming existing methods on benchmark datasets.

Contribution

The paper introduces SQUID, a new unsupervised anomaly detection approach that models anatomical structures with space-aware memory queues and creates a new dataset for evaluation.

Findings

SQUID surpasses 13 state-of-the-art methods by at least 5 AUC points.

It effectively identifies unseen or modified patterns in radiography images.

The new DigitAnatomy dataset facilitates development and evaluation of anomaly detection methods.

Abstract

Radiography imaging protocols focus on particular body regions, therefore producing images of great similarity and yielding recurrent anatomical structures across patients. To exploit this structured information, we propose the use of Space-aware Memory Queues for In-painting and Detecting anomalies from radiography images (abbreviated as SQUID). We show that SQUID can taxonomize the ingrained anatomical structures into recurrent patterns; and in the inference, it can identify anomalies (unseen/modified patterns) in the image. SQUID surpasses 13 state-of-the-art methods in unsupervised anomaly detection by at least 5 points on two chest X-ray benchmark datasets measured by the Area Under the Curve (AUC). Additionally, we have created a new dataset (DigitAnatomy), which synthesizes the spatial correlation and consistent shape in chest anatomy. We hope DigitAnatomy can prompt the development, evaluation, and interpretability of anomaly detection methods.