<p>Brain tumors are severe neurological disorders resulting from the abnormal proliferation of brain tissue cells, posing a significant threat to human health. Early detection and accurate localization are critical for effective diagnosis and treatment. However, existing approaches often face difficulties in achieving both high accuracy and computational efficiency, which limits their applicability in resource-constrained clinical environments. To address these challenges, this paper proposes a Multi-scale and Lightweight Synergistic YOLO model (MLS-YOLO) for brain tumor detection. The proposed model integrates multiple specialized modules to enhance feature representation and detection efficiency. Specifically, the MSRA module strengthens multi-scale feature extraction while suppressing redundant information; the DBSADown module effectively reduces feature map resolution while preserving informative features; and the SPPFE module leverages multi-scale pooling and an attention mechanism to improve detection performance for complex backgrounds and small targets. Extensive experiments conducted on the Brain Tumor Detection and Br35H datasets demonstrate that MLS-YOLO achieves improved detection performance while maintaining a lightweight design. Compared with YOLOv11n, MLS-YOLO improves mAP@0.5 by 2.2% and 2.0%, and mAP@0.5:0.95 by 1.3% and 2.6% on the two datasets, respectively. Meanwhile, the model reduces the number of parameters by 8.5% and achieves competitive inference efficiency (232.87 FPS and 214.33 FPS on the two datasets) under the same hardware conditions. These results indicate that MLS-YOLO achieves a favorable balance between detection accuracy and model efficiency, highlighting its potential for practical deployment in real-time and resource-limited clinical scenarios.</p>