Faculty, Staff and Student Publications

Language

English

Publication Date

1-1-2026

Journal

AMIA Summits on Translational Science Proceedings

PMID

42317829

PMCID

PMC13274361

PubMedCentral® Posted Date

6-1-2026

PubMedCentral® Full Text Version

Post-print

Abstract

Prompt-aware image segmentation has introduced new flexibility by allowing models to incorporate user-provided cues, such as points, boxes, or text prompts, to guide object delineation across diverse imaging domains. This flexibility introduces a fundamental challenge: ensuring segmentation results remain reproducible when prompts are altered. Minor variations in prompts produce substantially different outputs, raising concerns about the reliability and scientific validity of research and clinical applications. This paper conducts a structured review of emerging Prompt Instability Indices (PIIs), quantitative measures developed to assess the reproducibility of prompt-driven segmentation models. A systematic search of PubMed and arXiv (2020-2025) identified relevant studies, analyzed based on the definition, and operationalized prompt stability. This preliminary study suggests a taxonomy of PII methods, highlighting their respective advantages, limitations, and interpretive value. Integrating reproducibility metrics such as PIIs alongside traditional accuracy measures is essential for establishing consistent, transparent, and trustworthy performance assessments in next-generation interactive segmentation systems.

Published Open-Access

yes

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.