Lightweight Brain Tumor Detection in MRI Based on YOLO11 with Efficient Multi-Scale Attention and Wise-IoU Optimization
Traitement du Signal · 30.04.2026 · 43 (2) · 629-639
DOI: 10.18280/ts.430206 ↗Kurz gefasst
Fachartikel in der Zeitschrift Traitement du Signal, online erschienen am 30.04.2026. Verfasst von Ziyi Wang und 6 weiteren Autor:innen. Gefördert von: The Central Government's Guidance Fund forLocal Science and Technology Development, The Higher Education Research Project of Hebei Province, The Hebei Province Doctoral Student Innovation Capability Training Fund.
Abstract (Originalsprache)
Aiming at the clinical pain points in brain tumor medical image detection, such as sample category imbalance, significant tumor scale differences, inaccurate localization of lowquality samples and difficult model deployment, this paper proposes an improved brain tumor detection algorithm based on YOLO11.Firstly, a category balance strategy and combined data augmentation technology are adopted to solve the problem of uneven sample distribution, and improve the model's detection ability and robustness for tiny lesions.Secondly, an Efficient Multi-scale Attention (EMA) module is introduced, which accurately captures multi-scale tumor features without dimensionality reduction through feature grouping, parallel sub-networks and cross-spatial learning mechanism, reducing the loss of detailed information.Then, the Wise-IoU loss function is used to optimize gradient allocation through a dynamic non-monotonic focusing mechanism, weaken the interference of low-quality samples on bounding box regression, and improve localization accuracy.Finally, combined with the ShuffleNetV2 lightweight architecture, the number of model parameters and computation are reduced through grouped convolution and depthwise separable convolution technologies.The experimental results show that the improved algorithm achieves a precision of 96.3%, a recall of 95.4% and an mAP@50 of 96.7%.Meanwhile, the number of parameters is reduced by 41% and the computation is reduced by 43%.Lightweight deployment is realized on the premise of ensuring detection accuracy, which is highly adaptable to clinical auxiliary diagnosis and edge devices in primary medical institutions, and provides technical support for early accurate screening of brain tumors.
Förderung
- The Central Government's Guidance Fund forLocal Science and Technology Development
- The Higher Education Research Project of Hebei Province
- The Hebei Province Doctoral Student Innovation Capability Training Fund
Zitieren
APA
Wang, Z., Lv, R., Zhang, X., Song, Y., Meng, X., Liu, Z., & Jin, Q. (2026). Lightweight Brain Tumor Detection in MRI Based on YOLO11 with Efficient Multi-Scale Attention and Wise-IoU Optimization. Traitement du Signal, 43(2), 629-639. https://doi.org/10.18280/ts.430206
Vancouver
Wang Z, Lv R, Zhang X, Song Y, Meng X, Liu Z, et al. Lightweight Brain Tumor Detection in MRI Based on YOLO11 with Efficient Multi-Scale Attention and Wise-IoU Optimization. Traitement du Signal. 2026;43(2):629-639. doi:10.18280/ts.430206
BibTeX
@article{Wang2026,
author = {Wang, Ziyi and Lv, Ruobing and Zhang, Xu and Song, Yufei and Meng, Xi and Liu, Zhiguo and Jin, Qingyong},
title = {Lightweight Brain Tumor Detection in MRI Based on YOLO11 with Efficient Multi-Scale Attention and Wise-IoU Optimization},
journal = {Traitement du Signal},
year = {2026},
volume = {43},
number = {2},
pages = {629--639},
doi = {10.18280/ts.430206}
}
EUVIMED