This study explores the application of advanced detection techniques for anomaly detection in medical imaging. We formulate detection problem across various types of medical images and investigate the use of a modern object detection framework to address this challenge. Medical datasets originally designed for segmentation tasks are adapted to support effective tumor or anomaly detection, aiding computer-aided diagnosis by enabling malignancy recognition through bounding box annotations. The YOLOv8 detection framework is employed, integrated with various feature extraction backbones, and its detection performance is evaluated across different medical datasets. Additionally, a comparative analysis of tumor detection accuracy is conducted across different detection frameworks and datasets, highlighting the efficacy of recent approaches. Our experiments utilize a range of detection frameworks, including anchorless methods (e.g., CenterNet), two-stage methods (e.g., Faster R-CNN), and one-stage methods (e.g., YOLOv8). Notably, YOLOv8 demonstrates significant performance improvements across different datasets.