Foundation Models for Medical Image Segmentation Across Centers

Foundation models provide reusable representations, promptable interfaces and volumetric priors for medical image segmentation. Translating these capabilities into reliable performance at a new hospital requires attention to imaging variation, task definitions and the information available for adaptation. This critical narrative review focuses on organ and lesion segmentation in computed tomography and magnetic resonance imaging. It examines three connected questions: how transferable segmentation capabilities are formed, how adaptation mechanisms address differences between centers, and which external results support practical model selection. Pretraining objectives, dense prediction interfaces, support examples and spatial modeling are compared as complementary sources of capability. Transfer strategies are then organized by access to target images, training and validation labels, and case-specific interaction, with separate consideration of parameter, representation and input optimization at test time. Evidence from center-held-out studies, external disease cohorts and human annotation experiments shows that appropriate initialization, task adaptation and interaction can yield useful gains. Their interpretation depends on the comparator, the information supplied and the unit of evaluation. Accuracy, difficult cases, prompting requirements and the cost of the complete workflow therefore need to be considered together. An information-conditional framework links changes between centers to capability requirements, available resources, adaptation mechanisms and external benefit. This framework supports study design, model comparison and annotation resource allocation for deployment at new centers.

Authors

Institutions

Publication Details

Journal
X-Disciplinarity
Published
2026-11-18
DOI
https://doi.org/10.63174/xdi.lxhj8575
Primary Topic
Medical Image Segmentation Techniques
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Foundation Models for Medical Image Segmentation Across Centers

Zhicheng Zhang, Ruizhe Shi, Lele Zhao, Zhen Liu
3 links
X-Disciplinarity
Medical Image Segmentation Techniques
article

Foundation Models for Medical Image Segmentation Across Centers

Zhicheng Zhang, Ruizhe Shi, Lele Zhao, Zhen Liu
article en

Abstract

Foundation models provide reusable representations, promptable interfaces and volumetric priors for medical image segmentation. Translating these capabilities into reliable performance at a new hospital requires attention to imaging variation, task definitions and the information available for adaptation. This critical narrative review focuses on organ and lesion segmentation in computed tomography and magnetic resonance imaging. It examines three connected questions: how transferable segmentation capabilities are formed, how adaptation mechanisms address differences between centers, and which external results support practical model selection. Pretraining objectives, dense prediction interfaces, support examples and spatial modeling are compared as complementary sources of capability. Transfer strategies are then organized by access to target images, training and validation labels, and case-specific interaction, with separate consideration of parameter, representation and input optimization at test time. Evidence from center-held-out studies, external disease cohorts and human annotation experiments shows that appropriate initialization, task adaptation and interaction can yield useful gains. Their interpretation depends on the comparator, the information supplied and the unit of evaluation. Accuracy, difficult cases, prompting requirements and the cost of the complete workflow therefore need to be considered together. An information-conditional framework links changes between centers to capability requirements, available resources, adaptation mechanisms and external benefit. This framework supports study design, model comparison and annotation resource allocation for deployment at new centers.

X-DisciplinarityVol. 2(5)
Shandong Jiaotong University (CN), Shangdong Agriculture and Engineering University (CN), Shandong Huayu University of Technology (CN)
Shandong Agriculture and Engineering University
Good health and well-being
Medical Image Segmentation Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon