FM-OSD: Foundation Model-Enabled One-Shot Detection of Anatomical Landmarks

TL;DR

This paper introduces FM-OSD, a novel foundation model-based framework for one-shot anatomical landmark detection in medical images that requires only a single template image and no additional unlabeled data, achieving superior accuracy.

Contribution

The paper presents the first foundation model-enabled one-shot landmark detection method that eliminates the need for extensive unlabeled data by leveraging frozen visual foundation models and a novel bidirectional matching strategy.

Findings

Outperforms state-of-the-art one-shot landmark detection methods.

Effective with only a single template image and no extra unlabeled data.

Validated on two public anatomical landmark datasets.

Abstract

One-shot detection of anatomical landmarks is gaining significant attention for its efficiency in using minimal labeled data to produce promising results. However, the success of current methods heavily relies on the employment of extensive unlabeled data to pre-train an effective feature extractor, which limits their applicability in scenarios where a substantial amount of unlabeled data is unavailable. In this paper, we propose the first foundation model-enabled one-shot landmark detection (FM-OSD) framework for accurate landmark detection in medical images by utilizing solely a single template image without any additional unlabeled data. Specifically, we use the frozen image encoder of visual foundation models as the feature extractor, and introduce dual-branch global and local feature decoders to increase the resolution of extracted features in a coarse to fine manner. The introduced feature decoders are efficiently trained with a distance-aware similarity learning loss to incorporate domain knowledge from the single template image. Moreover, a novel bidirectional matching strategy is developed to improve both robustness and accuracy of landmark detection in the case of scattered similarity map obtained by foundation models. We validate our method on two public anatomical landmark detection datasets. By using solely a single template image, our method demonstrates significant superiority over strong state-of-the-art one-shot landmark detection methods.