| Title |
RAG⁴-Unet: an approach for recognition and segmentation of brain tumor in MRI scans |
| Authors |
Hamza, Ameer ; Damaševičius, Robertas |
| DOI |
10.15439/2025F7399 |
| ISBN |
9788397329195 |
| eISBN |
9788397329188 |
| Full Text |
|
| Is Part of |
Annals of computer science and information systems: position papers of the 20th conference on computer science and intelligence systems (FedCSIS), 14-17 September 2025, Kraków, Poland / M. Bolanowski, M. Ganzha, L. Maciaszek, M. Paprzycki, D. Ślęzak (eds.).. Warsaw : Polish Information Processing Society, 2025. vol. 44, p. 41-48.. ISSN 2300-5963. ISBN 9788397329195. eISBN 9788397329188 |
| Keywords [eng] |
Tumor Segmentation ; Residual Attention Gated ; Unet ; Yolo11 ; Attention Maps |
| Abstract [eng] |
We propose a novel U-net architecture, RAG${}^4$-Unet, based on residual attention gated for brain tumor segmentation, Swin transformer for classification task, and Yolo11 for tumor detection. For the experiments, the Figshare dataset is employed and the proposed architecture achieved 91.37\% Dice for tumor segmentation task, and Swin transformer achieved 91.74\% classification accuracy. The Yolo11 gained 89.6\% of detection precision. Comparative evaluation with the SOTA techniques reveals that the proposed architecture outperformed the existing methods and Yolo11. The proposed architecture improved the tumor boundary detection, making it a promising solution for brain tumor recognition and segmentation. |
| Published |
Warsaw : Polish Information Processing Society, 2025 |
| Type |
Conference paper |
| Language |
English |
| Publication date |
2025 |
| CC license |
|
