EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Towards Accurate Brain Tumour Segmentation: A Deep Learning Approach Using U-Net Variants

Maria Francisca Ferreira, Sofia Rosa, Pedro J.S.B. Caridade, Verónica Vasconcelos

Procedia Computer Science · 2026 · 278 · 1398-1403

DOI: 10.1016/j.procs.2026.03.125 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Procedia Computer Science, 2026. Verfasst von Maria Francisca Ferreira und 3 weiteren Autor:innen. Die Arbeit nennt 26 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Brain tumour segmentation in Magnetic Resonance Imaging (MRI) is essential for diagnosis and treatment planning in central nervous system disorders. In this paper, the deep learning models U-Net and U-Net++ are compared using a public dataset of brain MRI images and corresponding tumour masks. Both models were trained after undergoing pre-processing steps, including normalisation, resizing, and data augmentation. The U-Net achieved a high accuracy of 99.82% and a Dice Coefficient of 96.71% when trained with augmented data. U-Net++ also benefited from data augmentation, with improvements in all metrics, particularly in recall; however, it still achieved lower overall performance. These results highlight the effectiveness of both models and the impact of data augmentation in improving medical image segmentation. The simpler architecture of U-Net, combined with a well-designed augmentation pipeline and Dice-based loss function, facilitated better generalisation in this specific context.

Zitieren

APA

Ferreira, M. F., Rosa, S., Caridade, P. J., & Vasconcelos, V. (2026). Towards Accurate Brain Tumour Segmentation: A Deep Learning Approach Using U-Net Variants. Procedia Computer Science, 278, 1398-1403. https://doi.org/10.1016/j.procs.2026.03.125

Vancouver

Ferreira MF, Rosa S, Caridade PJ, Vasconcelos V. Towards Accurate Brain Tumour Segmentation: A Deep Learning Approach Using U-Net Variants. Procedia Computer Science. 2026;278:1398-1403. doi:10.1016/j.procs.2026.03.125

BibTeX
@article{Ferreira2026,
  author = {Ferreira, Maria Francisca and Rosa, Sofia and Caridade, Pedro J.S.B. and Vasconcelos, Verónica},
  title = {Towards Accurate Brain Tumour Segmentation: A Deep Learning Approach Using U-Net Variants},
  journal = {Procedia Computer Science},
  year = {2026},
  volume = {278},
  pages = {1398--1403},
  doi = {10.1016/j.procs.2026.03.125}
}
Kostenlos starten →