EUVIMEDEUVIMED entdecken
Crossref

Multi-Head Automated Segmentation by Incorporating Detection Head into the Contextual Layer Neural Network

Edwin Kys, Febian Febian

Open Access Journal of Applied Science and Technology · 10.02.2026 · 4 (1) · 01-07

DOI: 10.33140/oajast.04.01.10 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Open Access Journal of Applied Science and Technology, online erschienen am 10.02.2026. Verfasst von Edwin Kys und 1 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Deep-learning–based auto-segmentation is increasingly used in radiotherapy, but conventional models often produce anatomically implausible false positives, or “hallucinations,” in slices lacking target structures. We propose a gated multi-head Transformer architecture based on Swin U-Net, augmented with inter-slice context integration and a parallel detection head, which jointly performs slice-level structure detection via a multi-layer perceptron and pixel-level segmentation through a context-enhanced stream. Detection outputs gate the segmentation predictions to suppress false positives in anatomically invalid slices, and training uses slice-wise Tversky loss to address class imbalance. Experiments on the ProstateAnatomical-Edge-Cases dataset from The Cancer Imaging Archive demonstrate that the gated model substantially outperforms a non-gated segmentation-only baseline, achieving a mean Dice loss of 0.013 ± 0.036 versus 0.732 ± 0.314, with detection probabilities strongly correlated with anatomical presence, effectively eliminating spurious segmentations. In contrast, the non-gated model exhibited higher variability and persistent false positives across all slices. These results indicate that detection-based gating enhances robustness and anatomical plausibility in automated segmentation applications, reducing hallucinated predictions without compromising segmentation quality in valid slices, and offers a promising approach for improving the reliability of clinical radiotherapy auto-contouring workflows.

Zitieren

APA

Kys, E., & Febian, F. (2026). Multi-Head Automated Segmentation by Incorporating Detection Head into the Contextual Layer Neural Network. Open Access Journal of Applied Science and Technology, 4(1), 01-07. https://doi.org/10.33140/oajast.04.01.10

Vancouver

Kys E, Febian F. Multi-Head Automated Segmentation by Incorporating Detection Head into the Contextual Layer Neural Network. Open Access Journal of Applied Science and Technology. 2026;4(1):01-07. doi:10.33140/oajast.04.01.10

BibTeX
@article{Kys2026,
  author = {Kys, Edwin and Febian, Febian},
  title = {Multi-Head Automated Segmentation by Incorporating Detection Head into the Contextual Layer Neural Network},
  journal = {Open Access Journal of Applied Science and Technology},
  year = {2026},
  volume = {4},
  number = {1},
  pages = {01--07},
  doi = {10.33140/oajast.04.01.10}
}
Kostenlos starten →