Abstarct: Tumor detection is a critical task in medical imaging, aiming to identify abnormal growths with high accuracy and speed. Early and precise localization of tumors significantly improves treatment planning and patient outcomes. Deep learning-baxsed object detection models, especially those leveraging convolutional neural networks, have shown great promise in medical diagnostics. Among them, the YOLO family of real-time detectors offers a strong balance between speed and accuracy. In this study, we propose two enhanced versions of the YOLOv8 architecture, each integrated with a different attention mechanism to boost performance in tumor detection tasks. The first variant incorporates the Convolutional Block Attention Module (CBAM), which sequentially applies channel and spatial attention to refine feature representations. The second variant employs the Efficient Channel Attention (ECA) mechanism, which introduces lightweight yet effective channel attention by avoiding dimensionality reduction and capturing local cross-channel interactions. Both attention modules were integrated into the nech of YOLOv8 to enhance its focus on tumor-relevant regions in medical images. Experimental results on tumor datasets demonstrate that the proposed attention-augmented YOLOv8 models significantly outperform the baxseline in terms of precision, recall, and mean average precision . The inclusion of CBAM and ECA enhances the network's ability to differentiate between tumor and non-tumor regions, leading to more reliable detections. These improvements indicate the effectiveness of attention mechanisms in boosting the diagnostic performance of object detection models in medical imaging.