Deep learning is a rapidly advancing field, and computer vision techniques such as image segmentation and object detection have found extensive applications across various domains. In medical images, computer vision technology proves effective in handling large volumes of medical imaging data, thereby improving screening efficiency. In the realm of practical medical image applications, object detection has not gained the same level of popularity as image segmentation. Even though some object detection algorithms have been employed for lesion detection, the majority are single-stage detection algorithms, primarily based on the YOLO series[1–4]. However, the emergence of Transformers[2] appears to be altering this landscape. Leveraging the outstanding performance of Transformers, we propose the RPC-DETR model based on DETR[6–8], further exploring the potential of Transformers in lesion screening detection. We conducted experiments with the RPC-DETR model on the publicly available brain tumor dataset Br35H, minimizing the model's parameter count and reducing its complexity. In our experiments, RPC-DETR achieved high accuracy with only 14 million parameters. In summary, we have achieved greater accuracy in brain tumor detection by employing a more lightweight model.