AutoSeg -- Steering the Inductive Biases for Automatic Pathology Segmentation

TL;DR

AutoSeg introduces a novel approach that generates diverse artificial anomalies to improve the detection and segmentation of unseen pathologies in medical imaging, outperforming existing methods on chest X-ray datasets.

Contribution

The paper presents AutoSeg, a method that explicitly steers inductive biases to generate artificial anomalies, enhancing pathology segmentation and detection in medical images.

Findings

AutoSeg accurately segments unseen artificial anomalies.

Outperforms existing methods on Chest X-ray datasets.

Effective in real-world out-of-distribution detection tasks.

Abstract

In medical imaging, un-, semi-, or self-supervised pathology detection is often approached with anomaly- or out-of-distribution detection methods, whose inductive biases are not intentionally directed towards detecting pathologies, and are therefore sub-optimal for this task. To tackle this problem, we propose AutoSeg, an engine that can generate diverse artificial anomalies that resemble the properties of real-world pathologies. Our method can accurately segment unseen artificial anomalies and outperforms existing methods for pathology detection on a challenging real-world dataset of Chest X-ray images. We experimentally evaluate our method on the Medical Out-of-Distribution Analysis Challenge 2021.