SCMamba-SAM: transferring knowledge from natural images to medical tasks
6 affiliations
- 1Medical School, Tianjin University
- 2The First Affiliated Hospital of Hebei North University
- 3Graduate School of Tianjin Medical University
- 4College of Computer Science and Artificial Intelligence, Fudan University
- 5Tencent Jarvis Lab
- 6Jilin Normal University
Abstract
Background and objective: The Segment Anything Model (SAM) has drawn significant attention because of its powerful performance in generating high-quality objectmasks. Despite its superior performance in handling natural images, SAM exhibits significant performance degradation when applied to medical images due to the substantial domain gap between natural and medical images. In addition, the specific characteristics of medical images, such as strong heterogeneity, low contrast, and wide individual variation, further raise dificulties in segmenting medical images.
Methods: To bridge the semantic gap between natural and medical images, we propose an adaption framework, named SCMamba-SAM. It employs a foundational image encoder constructed by integrating the pre-trained encoder of SAM and Low-Rank Adaption technique. The Encoding Mamba Adapter adopts the spatial- and channel-Mamba blocks to capture task-related global features. Additionally, an enhanced mask decoder with deep supervision has been developed to eliminate the requirement for a prompt encoder, thereby transforming SAM into a fully automated and highly accurate image segmentation model.
Results: Extensive experiments were performed on six public lesion segmentation datasets in different imaging modalities and body parts. The proposed SCMamba-SAM not only achieved performance superior to that of SOTA methods, quantitatively and qualitatively, but also exhibited high generalization.
Conclusion: SCMamba-SAM holds promise as a useful tool for clinicians to diagnose and quantify lesions, providing important information for treatment decisions and surgery planning.
Related
- Two-Year Knowledge Retention After a Short Online Palliative Care Intervention for Medical Students: A Controlled Experimental Study
- Medical Students’ Knowledge, Awareness, and Behaviors Regarding Food, Dietary Supplement, and Drug Interactions Across Academic Years
- A Robust and Trustable Approach to Incorporate Medical Domain Knowledge in Machine Learning Models for Diabetic Retinopathy Screening Using Routine Lab Results
- Assessing theoretical knowledge related to surgical and clinical skills in Syrian medical students : A cross-sectional study.
- Nutrition Knowledge, Information Sources, and Self-Perceived Ability to Provide Nutrition Counseling Among Medical Students: A Cross-Sectional Study
BibTeX
@article{tian2026scmambasam,
title = {{SCMamba-SAM: transferring knowledge from natural images to medical tasks}},
author = {Ye Tian and Yufei Yuan and Lu Wang and Xinming Yang and Yifeng Wu and Xu Xu},
journal = {Frontiers in Neurology},
year = {2026},
month = oct,
doi = {10.3389/fneur.2026.1938737},
url = {https://www.frontiersin.org/articles/10.3389/fneur.2026.1938737/full},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - SCMamba-SAM: transferring knowledge from natural images to medical tasks
AU - Ye Tian
AU - Yufei Yuan
AU - Lu Wang
AU - Xinming Yang
AU - Yifeng Wu
AU - Xu Xu
PY - 2026
DA - 2026/10/15
JO - Frontiers in Neurology
DO - 10.3389/fneur.2026.1938737
UR - https://www.frontiersin.org/articles/10.3389/fneur.2026.1938737/full
AB - Background and objective: The Segment Anything Model (SAM) has drawn significant attention because of its powerful performance in generating high-quality objectmasks. Despite its superior performance in handling natural images, SAM exhibits significant performance degradation when applied to medical images due to the substantial domain gap between natural and medical images. In addition, the specific characteristics of medical images, such as strong heterogeneity, low contrast, and wide individual variation, further raise dificulties in segmenting medical images. Methods: To bridge the semantic gap between natural and medical images, we propose an adaption framework, named SCMamba-SAM. It employs a foundational image encoder constructed by integrating the pre-trained encoder of SAM and Low-Rank Adaption technique. The Encoding Mamba Adapter adopts the spatial- and channel-Mamba blocks to capture task-related global features. Additionally, an enhanced mask decoder with deep supervision has been developed to eliminate the requirement for a prompt encoder, thereby transforming SAM into a fully automated and highly accurate image segmentation model. Results: Extensive experiments were performed on six public lesion segmentation datasets in different imaging modalities and body parts. The proposed SCMamba-SAM not only achieved performance superior to that of SOTA methods, quantitatively and qualitatively, but also exhibited high generalization. Conclusion: SCMamba-SAM holds promise as a useful tool for clinicians to diagnose and quantify lesions, providing important information for treatment decisions and surgery planning.
SN - 1664-2295
ER -