From Pixels to Graphs: Deep Graph-Level Anomaly Detection on Dermoscopic Images

TL;DR

This paper systematically evaluates image-to-graph transformation methods for graph-level anomaly detection on dermoscopic images using GNNs, highlighting the impact of different features and supervision levels on detection performance.

Contribution

It provides a comprehensive comparison of segmentation, edge construction, and feature extraction strategies for GNN-based anomaly detection on dermoscopic images, filling a gap in prior research.

Findings

Color descriptors yield the best standalone performance.

Incorporating shape and texture features improves detection accuracy.

Supervised models achieve up to 0.914 AUC-ROC on dermoscopic images.

Abstract

Graph Neural Networks (GNNs) have emerged as a powerful approach for graph-based machine learning tasks. Previous work applied GNNs to image-derived graph representations for various downstream tasks such as classification or anomaly detection. These transformations include segmenting images, extracting features from segments, mapping them to nodes, and connecting them. However, to the best of our knowledge, no study has rigorously compared the effectiveness of the numerous potential image-to-graph transformation approaches for GNN-based graph-level anomaly detection (GLAD). In this study, we systematically evaluate the efficacy of multiple segmentation schemes, edge construction strategies, and node feature sets based on color, texture, and shape descriptors to produce suitable image-derived graph representations to perform graph-level anomaly detection. We conduct extensive experiments on dermoscopic images using state-of-the-art GLAD models, examining performance and efficiency in purely unsupervised, weakly supervised, and fully supervised regimes. Our findings reveal, for example, that color descriptors contribute the best standalone performance, while incorporating shape and texture features consistently enhances detection efficacy. In particular, our best unsupervised configuration using OCGTL achieves a competitive AUC-ROC score of up to 0.805 without relying on pretrained backbones like comparable image-based approaches. With the inclusion of sparse labels, the performance increases substantially to 0.872 and with full supervision to 0.914 AUC-ROC.