Superpixel-graph self-supervised pretraining with joint-embedding and contrastive objectives for wound image segmentation
Advances in Science and Technology Research Journal · 01.09.2026 · 20 (9) · 163-174
DOI: 10.12913/22998624/224196 ↗Kurz gefasst
Fachartikel in der Zeitschrift Advances in Science and Technology Research Journal, online erschienen am 01.09.2026. Verfasst von Bohdan Lukashchuk und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Annotated data remain a major problem and bottleneck for supervised learning approaches, especially in wound imaging. In this work, a self-supervised learning pretraining method was proposed that represents a wound image as a superpixel adjacency graph, generated using the simple linear iterative clustering algorithm and uses graph structure to define building blocks for calculating joint-embedding and contrastive learning objectives. Three objective variants were evaluated – contrastive, joint-embedding and their combination as a pretraining stage for further finetuning for the wound segmentation task. After self-supervised pretraining on 100 unlabelled images and supervised fine-tuning on only 30 labelled images using the contrastive objective, a Dice score of 0.266 was received on a 400-image test set, compared to 0.354 for a fully supervised U-Net trained on all 100 annotated images, thus having a gap of 0.088 in Dice score with more than three times less annotated data. The results sug - gest that superpixel-graph-based self-supervised learning is a promising pretraining strategy for wound analysis in settings with limited annotated data.
Zitieren
APA
Lukashchuk, B., & Farmaha, I. (2026). Superpixel-graph self-supervised pretraining with joint-embedding and contrastive objectives for wound image segmentation. Advances in Science and Technology Research Journal, 20(9), 163-174. https://doi.org/10.12913/22998624/224196
Vancouver
Lukashchuk B, Farmaha I. Superpixel-graph self-supervised pretraining with joint-embedding and contrastive objectives for wound image segmentation. Advances in Science and Technology Research Journal. 2026;20(9):163-174. doi:10.12913/22998624/224196
BibTeX
@article{Lukashchuk2026,
author = {Lukashchuk, Bohdan and Farmaha, Ihor},
title = {Superpixel-graph self-supervised pretraining with joint-embedding and contrastive objectives for wound image segmentation},
journal = {Advances in Science and Technology Research Journal},
year = {2026},
volume = {20},
number = {9},
pages = {163--174},
doi = {10.12913/22998624/224196}
}
EUVIMED