The study describes the ESA‑YOLOv5m model — a modified version of YOLOv5m supplemented with an Enhanced Spatial Attention (ESA) module — intended for the detection of brain tumors in MRI images. The ESA module is placed behind the Spatial Pyramid Pooling‑Fast (SPPF) layer and is intended to highlight diagnostically relevant regions and suppress background noise without a significant increase in computational complexity[1]. Experiments were performed on a Figshare dataset containing three tumor classes: glioma, meningioma, and pituitary[1]. The ESA-YOLOv5m model achieved a precision of 90%, a recall of 90% and a median average accuracy of mAP@0.5 of 91%[1]. An ablation study showed that placing the ESA module behind the SPPF gave the highest performance (mAP@0.5 = 0.91), while earlier insertion produced slightly lower results[1]. Five-fold cross-validation confirmed stable performance with mAP@0.5 = 0.910 ± 0.006[1]. Performance tests showed negligible overhead — less than 4.3% parameter increase — and average latency below 10ms per image[1]. The authors conclude that the framework provides a reliable and scalable solution for automated brain tumor detection suitable for clinical support systems and edge applications[1].