The Role of Attention in Multi-Sequence MRI Brain Tumor Segmentation using modified Swin UNETR
International Journal of Drug Delivery Technology · 10.06.2026 · 16 (47s)
DOI: 10.25258/ijddt.16.47s.151 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Drug Delivery Technology, online erschienen am 10.06.2026. Verfasst von Sonia F Panesar und 1 weiteren Autor:innen.
Abstract (Originalsprache)
Multiple sequences of MRI play a pivotal role in brain tumor diagnosis, treatment planning, and decision prediction. Due to inherent spatial heterogeneity among patients, accurate segmentation continuous to be a complex challenge. While convolutional neural networks (CNNs) like U-Net and Transformer-based models such as ViT and Swin UNETR have shown strong performance in capturing local and global context in computer vision tasks, their feature representations can further be enhanced. In our work, we explore the effectiveness of an auxiliary lightweight attention mechanisms in improving the segmentation capability of Swin UNETR. Two proposed integrations are investigated: Swin UNETR enhanced with Squeezeand-Excitation (SE) blocks, and Swin UNETR enhanced with the Convolutional Block Attention Module (CBAM). Evaluation results on the BraTS2020 dataset reveal that both auxilliary attention mechanism improve segmentation accuracy over the baseline model to a certain extent. The CBAM-enhanced model achieves superior performance across all metrics, with a Dice score of 0.8641, IoU of 0.7884, Recall of 0.9124, Hausdorff Distance of 3.2596, and ASSD of 0.2859, accompanied by increased model complexity. The results achieved conclude that optimal attention design and placement will remain an open research question. Future work will focus on developing adaptive and computationally efficient attention focussed strategies as well as assessing the generalizability across multi-center datasets, to provide a foundation for Federated Learning and clinical deployment.
Zitieren
APA
Panesar, S. F., & Ganatra, A. P. (2026). The Role of Attention in Multi-Sequence MRI Brain Tumor Segmentation using modified Swin UNETR. International Journal of Drug Delivery Technology, 16(47s). https://doi.org/10.25258/ijddt.16.47s.151
Vancouver
Panesar SF, Ganatra AP. The Role of Attention in Multi-Sequence MRI Brain Tumor Segmentation using modified Swin UNETR. International Journal of Drug Delivery Technology. 2026;16(47s). doi:10.25258/ijddt.16.47s.151
BibTeX
@article{Panesar2026,
author = {Panesar, Sonia F and Ganatra, Amit P},
title = {The Role of Attention in Multi-Sequence MRI Brain Tumor Segmentation using modified Swin UNETR},
journal = {International Journal of Drug Delivery Technology},
year = {2026},
volume = {16},
number = {47s},
doi = {10.25258/ijddt.16.47s.151}
}
EUVIMED