EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Improving Pinworm Eggs Segmentation With Stacked Transfer Learning-Based U-Net Ensembles

Emre Özbilge, Mehmet Karlitaş, Emrah Güler, Ebru Ozbilge, Hüseyın Kaya Süer

IEEE Access · 2026 · 14 · 67072-67087

DOI: 10.1109/access.2026.3689421 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Emre Özbilge und 4 weiteren Autor:innen. Die Arbeit nennt 31 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Human parasites negatively impact health and reduce the quality of life worldwide. Therefore, preventing the spread of parasites and treating infections are essential for public health. Pinworm parasites are among the most common human parasites and directly affect sleep quality. It has a worldwide distribution and is the leading cause of helminthiasis in the United States, Western Europe, and Oceania. Enterobiasis, caused by pinworm parasites, significantly impairs sleep quality and restfulness, particularly in children. This study aimed to semantically segment pinworm parasite eggs in microscopic images using deep learning techniques. The proposed approach assists clinical experts by enabling more accurate and consistent detection of parasite eggs, thereby overcoming the limitations of visual inspection under a microscope. Improved identification supports timely treatment and enhances patients’ overall well-being. We investigated a baseline U-Net model and several transfer-learning variants in which the encoder was replaced with pretrained convolutional networks, including VGG19, EfficientNetV2M, MobileNetV2, ResNet50V2, and DenseNet201. In addition, a stacked ensemble framework is proposed, in which the predictions of multiple base models are combined using a meta U-Net. The experimental results showed that DenseNet201 +U-Net achieved the strongest standalone performance, with a Dice coefficient of 0.9135. Based on this, the proposed stacking ensemble further improves the segmentation accuracy. In particular, the ensemble combining DenseNet201, EfficientNetV2M, and MobileNetV2 with a U-Net meta-learner achieved the best overall performance, with a Dice coefficient of 0.9257. These findings confirm that the selective stacking of complementary architectures provides a robust and effective solution for pinworm-egg segmentation in microscopic images.

Zitieren

APA

Özbilge, E., Karlitaş, M., Güler, E., Ozbilge, E., & Süer, H. K. (2026). Improving Pinworm Eggs Segmentation With Stacked Transfer Learning-Based U-Net Ensembles. IEEE Access, 14, 67072-67087. https://doi.org/10.1109/access.2026.3689421

Vancouver

Özbilge E, Karlitaş M, Güler E, Ozbilge E, Süer HK. Improving Pinworm Eggs Segmentation With Stacked Transfer Learning-Based U-Net Ensembles. IEEE Access. 2026;14:67072-67087. doi:10.1109/access.2026.3689421

BibTeX
@article{zbilge2026,
  author = {Özbilge, Emre and Karlitaş, Mehmet and Güler, Emrah and Ozbilge, Ebru and Süer, Hüseyın Kaya},
  title = {Improving Pinworm Eggs Segmentation With Stacked Transfer Learning-Based U-Net Ensembles},
  journal = {IEEE Access},
  year = {2026},
  volume = {14},
  pages = {67072--67087},
  doi = {10.1109/access.2026.3689421}
}
Kostenlos starten →