SAM Fewshot Finetuning for Anatomical Segmentation in Medical Images

TL;DR

This paper introduces a simple yet effective few-shot fine-tuning method for adapting SAM to medical image segmentation, significantly reducing labeling effort while maintaining high accuracy across multiple datasets.

Contribution

The authors reformulate SAM's mask decoder to utilize few-shot embeddings as prompts, enabling efficient adaptation to medical segmentation tasks with minimal labeled data.

Findings

Achieves about 50% improvement in IoU over point prompt SAM.

Performs comparably to fully supervised methods with much less labeled data.

Validated on four datasets across six anatomical segmentation tasks.

Abstract

We propose a straightforward yet highly effective few-shot fine-tuning strategy for adapting the Segment Anything (SAM) to anatomical segmentation tasks in medical images. Our novel approach revolves around reformulating the mask decoder within SAM, leveraging few-shot embeddings derived from a limited set of labeled images (few-shot collection) as prompts for querying anatomical objects captured in image embeddings. This innovative reformulation greatly reduces the need for time-consuming online user interactions for labeling volumetric images, such as exhaustively marking points and bounding boxes to provide prompts slice by slice. With our method, users can manually segment a few 2D slices offline, and the embeddings of these annotated image regions serve as effective prompts for online segmentation tasks. Our method prioritizes the efficiency of the fine-tuning process by exclusively training the mask decoder through caching mechanisms while keeping the image encoder frozen. Importantly, this approach is not limited to volumetric medical images, but can generically be applied to any 2D/3D segmentation task. To thoroughly evaluate our method, we conducted extensive validation on four datasets, covering six anatomical segmentation tasks across two modalities. Furthermore, we conducted a comparative analysis of different prompting options within SAM and the fully-supervised nnU-Net. The results demonstrate the superior performance of our method compared to SAM employing only point prompts (approximately 50% improvement in IoU) and performs on-par with fully supervised methods whilst reducing the requirement of labeled data by at least an order of magnitude.