SAIF: A Stability-Aware Inference Framework for Medical Image Segmentation with Segment Anything Model

TL;DR

SAIF is a training-free inference framework that enhances the stability and accuracy of medical image segmentation using the Segment Anything Model by modeling and filtering uncertainty during inference.

Contribution

SAIF introduces a novel, plug-and-play method that explicitly models prompt and threshold uncertainty to improve segmentation robustness without retraining.

Findings

SAIF improves segmentation accuracy across multiple datasets.

SAIF enhances robustness against localization errors and threshold variations.

SAIF achieves state-of-the-art performance without retraining.

Abstract

Segment Anything Model (SAM) enable scalable medical image segmentation but suffer from inference-time instability when deployed as a frozen backbone. In practice, bounding-box prompts often contain localization errors, and fixed threshold binarization introduces additional decision uncertainty. These factors jointly cause high prediction variance, especially near object boundaries, degrading reliability. We propose the Stability-Aware Inference Framework (SAIF), a training-free and plug-and-play inference framework that improves robustness by explicitly modeling prompt and threshold uncertainty. SAIF constructs a joint uncertainty space via structured box perturbations and threshold variations, evaluates each hypothesis using decision stability and boundary consistency, and introduces a stability-consistency score to filter unstable candidates and perform stability-weighted fusion in probability space. Experiments on Synapse, CVC-ClinicDB, Kvasir-SEG, and CVC-300 demonstrate that SAIF consistently improves segmentation accuracy and robustness, achieving state-of-the-art performance without retraining or architectural modification. Our anonymous code is released at https://anonymous.4open.science/r/SAIF.