Position Regression for Unsupervised Anomaly Detection

TL;DR

This paper introduces a novel coordinate regression method for unsupervised anomaly detection in 3D medical images, which localizes anomalies by estimating patch positions and requires less memory than reconstruction-based methods.

Contribution

The paper proposes a new coordinate regression approach for anomaly detection that is effective and memory-efficient for large 3D medical images, trained solely on healthy data.

Findings

Accurate detection of intracranial hemorrhages and fractures.

Requires less memory than reconstruction-based methods.

Effective in localizing anomalies in 3D CT volumes.

Abstract

In recent years, anomaly detection has become an essential field in medical image analysis. Most current anomaly detection methods for medical images are based on image reconstruction. In this work, we propose a novel anomaly detection approach based on coordinate regression. Our method estimates the position of patches within a volume, and is trained only on data of healthy subjects. During inference, we can detect and localize anomalies by considering the error of the position estimate of a given patch. We apply our method to 3D CT volumes and evaluate it on patients with intracranial haemorrhages and cranial fractures. The results show that our method performs well in detecting these anomalies. Furthermore, we show that our method requires less memory than comparable approaches that involve image reconstruction. This is highly relevant for processing large 3D volumes, for instance, CT or MRI scans.