Cross-Domain Robust Pruning for Polyp Segmentation: Multi-Encoder Feature Fusion Beats Single-Encoder Baselines
Bioengineering · 29.06.2026 · 13 (7) · 759
DOI: 10.3390/bioengineering13070759 ↗Kurz gefasst
Fachartikel in der Zeitschrift Bioengineering, online erschienen am 29.06.2026. Verfasst von Chia-Pei Tang und 4 weiteren Autor:innen. Beteiligte Einrichtung: Division of Gastroenterology, Department of Internal Medicine, Dalin Tzu Chi Hospital, Buddhist Tzu Chi Medical Foundation, Chiayi City 622401, Taiwa…. Die Arbeit nennt 30 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Science and Technology Council, Dalin Tzu Chi Hospital, Buddhist Tzu Chi Medical Foundation, National Science and Technology Council (NSTC) of Taiwan.
Abstract (Originalsprache)
Medical image segmentation requires dense pixel-level annotations, making large-scale dataset construction expensive and motivating research into data-efficient training. The main objective of this paper is to determine whether fusing two complementary pretrained image encoders into a single similarity space can make training-free dataset pruning robust across heterogeneous polyp segmentation domains and to quantify that robustness against a comprehensive panel of baselines. To achieve this, we propose Multi-Encoder Diverse Pruning (MEDP), a training-free dataset-pruning method. MEDP fuses features from an ImageNet-pretrained ResNet-18 and a self-supervised DINOv2 ViT-S/14 into a single 896-D similarity space. It partitions the training pool via Louvain modularity maximization and selects per-community samples via maximal-marginal-relevance (MMR) ranking, which effectively balances eigenvector centrality with feature-space diversity. We benchmarked MEDP against 12 baselines at a 20% retention ratio across three polyp segmentation settings (Kvasir-SEG, CVC-ClinicDB, and a Combined cross-domain pool) using a standard 5-level UNet. Based on approximately 330 controlled training runs, the results demonstrate that MEDP achieves the highest mean test Dice of 0.7324 on the most challenging Combined cross-domain pool, significantly outperforming uniform random sampling (Cohen’s d = +1.79, paired Wilcoxon p = 0.002). Conversely, all hand-crafted structure-aware variants failed to outperform uniform random sampling. These findings confirm that combining multi-encoder features with MMR diversity provides a simple and effective strategy for improving robustness across heterogeneous medical imaging settings and that the choice of pretrained image encoder is the dominant factor in segmentation-aware pruning.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Division of Gastroenterology, Department of Internal Medicine, Dalin Tzu Chi Hospital, Buddhist Tzu Chi Medical Foundation, Chiayi City 622401, Taiwan.
- School of Medicine, Tzu Chi University, Hualien City 970374, Taiwan.
- Department of Management Information System, National Chiayi University, Chiayi City 600355, Taiwan.
- Institute of Information Security, National Tsing Hua University, Hsinchu City 600355, Taiwan.
- Department of Information Management, National Yunlin University of Science and Technology, Douliu City 64002, Taiwan.
- Department of Medical Research, Dalin Tzu Chi Hospital, Buddhist Tzu Chi Medical Foundation, Chiayi City 622401, Taiwan.
ORCID-Profile: Hong-Yi Chang
Förderung
- National Science and Technology Council
- Dalin Tzu Chi Hospital, Buddhist Tzu Chi Medical Foundation
- National Science and Technology Council (NSTC) of Taiwan
Zitieren
APA
Tang, C. P., Chang, H. Y., Chang, T. S., Chang, Y. C., & Cheng, C. H. (2026). Cross-Domain Robust Pruning for Polyp Segmentation: Multi-Encoder Feature Fusion Beats Single-Encoder Baselines. Bioengineering, 13(7), 759. https://doi.org/10.3390/bioengineering13070759
Vancouver
Tang CP, Chang HY, Chang TS, Chang YC, Cheng CH. Cross-Domain Robust Pruning for Polyp Segmentation: Multi-Encoder Feature Fusion Beats Single-Encoder Baselines. Bioengineering. 2026;13(7):759. doi:10.3390/bioengineering13070759
BibTeX
@article{Tang2026,
author = {Tang, Chia-Pei and Chang, Hong-Yi and Chang, Tzu-Shan and Chang, Yu-Chieh and Cheng, Chia-Hsin},
title = {Cross-Domain Robust Pruning for Polyp Segmentation: Multi-Encoder Feature Fusion Beats Single-Encoder Baselines},
journal = {Bioengineering},
year = {2026},
volume = {13},
number = {7},
pages = {759},
doi = {10.3390/bioengineering13070759}
}
EUVIMED