3DSAM-adapter: Holistic adaptation of SAM from 2D to 3D for promptable tumor segmentation

TL;DR

This paper introduces 3DSAM-adapter, a novel method that adapts the 2D SAM model for 3D medical tumor segmentation, achieving superior results with minimal retraining on volumetric data.

Contribution

The paper presents a holistic architecture modification and lightweight fine-tuning approach to transfer SAM from 2D to 3D medical imaging, maintaining pre-trained knowledge while enhancing segmentation performance.

Findings

Outperforms state-of-the-art models on 3 out of 4 tumor segmentation datasets.

Achieves up to 29.87% improvement in segmentation accuracy.

Requires minimal additional parameters for effective 3D adaptation.

Abstract

Despite that the segment anything model (SAM) achieved impressive results on general-purpose semantic segmentation with strong generalization ability on daily images, its demonstrated performance on medical image segmentation is less precise and not stable, especially when dealing with tumor segmentation tasks that involve objects of small sizes, irregular shapes, and low contrast. Notably, the original SAM architecture is designed for 2D natural images, therefore would not be able to extract the 3D spatial information from volumetric medical data effectively. In this paper, we propose a novel adaptation method for transferring SAM from 2D to 3D for promptable medical image segmentation. Through a holistically designed scheme for architecture modification, we transfer the SAM to support volumetric inputs while retaining the majority of its pre-trained parameters for reuse. The fine-tuning process is conducted in a parameter-efficient manner, wherein most of the pre-trained parameters remain frozen, and only a few lightweight spatial adapters are introduced and tuned. Regardless of the domain gap between natural and medical data and the disparity in the spatial arrangement between 2D and 3D, the transformer trained on natural images can effectively capture the spatial patterns present in volumetric medical images with only lightweight adaptations. We conduct experiments on four open-source tumor segmentation datasets, and with a single click prompt, our model can outperform domain state-of-the-art medical image segmentation models on 3 out of 4 tasks, specifically by 8.25%, 29.87%, and 10.11% for kidney tumor, pancreas tumor, colon cancer segmentation, and achieve similar performance for liver tumor segmentation. We also compare our adaptation method with existing popular adapters, and observed significant performance improvement on most datasets.