EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

DynU-Net: Dynamic uncertainty-aware multi-task U-Net for joint lesion segmentation and classification in medical imaging

Ngoc Ly Tran, Thi Thu Thuy Nguyen, Ba Hung Ngo, Tae Jong Choi

Journal of Computational Design and Engineering · 04.08.2026 · 13 (8) · 109-132

DOI: 10.1093/jcde/qwag069 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Computational Design and Engineering, online erschienen am 22.07.2026. Verfasst von Ngoc Ly Tran und 3 weiteren Autor:innen. Die Arbeit nennt 58 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Research Foundation of Korea.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Segmentation and classification of lesion is essential for computer-aided diagnosis in grayscale medical imaging, enabling precise localization and reliable categorization of various pathological conditions. However, conventional multi-task learning (MTL) frameworks often suffer from suboptimal task interactions and require manual loss weight tuning, limiting their effectiveness on challenging modalities such as ultrasound, magnetic resonance imaging (MRI), and computed tomography (CT). To address these limitations, we propose a Dynamic Uncertainty-aware Network (DynU-Net), a multi-task framework that adaptively balances segmentation and classification through learnable per-task uncertainty parameters. This mechanism eliminates the need for manual hyperparameter tuning while mitigating gradient imbalance during optimization. Extensive experiments on three public datasets, including Brain Stroke CT, Brain Tumor MRI, and BUSI (ultrasound), demonstrate that DynU-Net consistently outperforms both single-task and existing multi-task baselines. In particular, it achieves a Dice score of $88.90\%\pm 0.33\%$ and a macro-F1 score of $96.73\%\pm 0.74\%$ on the Brain Stroke CT dataset while maintaining favorable computational efficiency compared to competing architectures. These results highlight the effectiveness of dynamic uncertainty-aware optimization in enabling robust and balanced MTL for joint lesion segmentation and classification on public medical imaging benchmarks.

Förderung

  • National Research Foundation of Korea

Zitieren

APA

Tran, N. L., Nguyen, T. T. T., Ngo, B. H., & Choi, T. J. (2026). DynU-Net: Dynamic uncertainty-aware multi-task U-Net for joint lesion segmentation and classification in medical imaging. Journal of Computational Design and Engineering, 13(8), 109-132. https://doi.org/10.1093/jcde/qwag069

Vancouver

Tran NL, Nguyen TTT, Ngo BH, Choi TJ. DynU-Net: Dynamic uncertainty-aware multi-task U-Net for joint lesion segmentation and classification in medical imaging. Journal of Computational Design and Engineering. 2026;13(8):109-132. doi:10.1093/jcde/qwag069

BibTeX
@article{Tran2026,
  author = {Tran, Ngoc Ly and Nguyen, Thi Thu Thuy and Ngo, Ba Hung and Choi, Tae Jong},
  title = {DynU-Net: Dynamic uncertainty-aware multi-task U-Net for joint lesion segmentation and classification in medical imaging},
  journal = {Journal of Computational Design and Engineering},
  year = {2026},
  volume = {13},
  number = {8},
  pages = {109--132},
  doi = {10.1093/jcde/qwag069}
}
Kostenlos starten →