DINO-AD: Unsupervised Anomaly Detection with Frozen DINO-V3 Features

TL;DR

DINO-AD introduces a novel unsupervised anomaly detection method in medical imaging that leverages self-supervised DINO-V3 features, combining embedding similarity and clustering for accurate and interpretable localization.

Contribution

The paper presents a new framework using DINO-V3 features with embedding matching and foreground-aware clustering for improved anomaly detection without annotations.

Findings

Achieves up to 98.71 AUROC on medical datasets

Produces clearer and more accurate anomaly localization

Demonstrates robustness and generalizability through ablation studies

Abstract

Unsupervised anomaly detection (AD) in medical images aims to identify abnormal regions without relying on pixel-level annotations, which is crucial for scalable and label-efficient diagnostic systems. In this paper, we propose a novel anomaly detection framework based on DINO-V3 representations, termed DINO-AD, which leverages self-supervised visual features for precise and interpretable anomaly localization. Specifically, we introduce an embedding similarity matching strategy to select a semantically aligned support image and a foreground-aware K-means clustering module to model the distribution of normal features. Anomaly maps are then computed by comparing the query features with clustered normal embeddings through cosine similarity. Experimental results on both the Brain and Liver datasets demonstrate that our method achieves superior quantitative performance compared with state-of-the-art approaches, achieving AUROC scores of up to 98.71. Qualitative results further confirm that our framework produces clearer and more accurate anomaly localization. Extensive ablation studies validate the effectiveness of each proposed component, highlighting the robustness and generalizability of our approach.