This paper introduces the application of YOLOv9, the latest version of the You Only Look Once (YOLO) series, for fracture detection in pediatric wrist trauma X-ray images. The YOLOv9 model was trained on the GRAZPEDWRI-DX dataset, which contains 20,327 X-ray images of pediatric wrist trauma. To improve model performance, data augmentation techniques were used to extend the training set. Experimental results show that the YOLOv9 model achieved a higher mAP 50-95 value (43.73%) compared to the current state-of-the-art (SOTA) model (42.16%), with an improvement of 3.7%. The model's performance was evaluated on input image sizes of 640 and 1024, with YOLOv9-E achieving a mAP 50-95 of 43.73% when the input size was 1024, demonstrating SOTA performance. The YOLOv9 model was found to be effective in balancing inference time with fracture detection accuracy, making it suitable for deployment in web applications as a CAD system to assist surgeons in analyzing X-ray images. The model's ability to accurately predict "fracture", "metal", and "text" classes was high, with YOLOv9 achieving a 94.8% accuracy in "metal" class prediction. However, the performance in predicting bone anomalies and soft tissue classes was poorer, with YOLOv9 achieving only 11.3% and 28.2% accuracy, respectively. The paper also discusses the limitations of the current dataset and suggests future research directions to enhance model performance by improving the models with additional information on "bone anomaly" and "soft tissue" classes. The trained weight files are available on the authors' GitHub for further use. The study demonstrates that the YOLOv9 model can serve as a CAD system to assist radiologists and surgeons in interpreting X-ray images, particularly in pediatric wrist fracture detection.This paper introduces the application of YOLOv9, the latest version of the You Only Look Once (YOLO) series, for fracture detection in pediatric wrist trauma X-ray images. The YOLOv9 model was trained on the GRAZPEDWRI-DX dataset, which contains 20,327 X-ray images of pediatric wrist trauma. To improve model performance, data augmentation techniques were used to extend the training set. Experimental results show that the YOLOv9 model achieved a higher mAP 50-95 value (43.73%) compared to the current state-of-the-art (SOTA) model (42.16%), with an improvement of 3.7%. The model's performance was evaluated on input image sizes of 640 and 1024, with YOLOv9-E achieving a mAP 50-95 of 43.73% when the input size was 1024, demonstrating SOTA performance. The YOLOv9 model was found to be effective in balancing inference time with fracture detection accuracy, making it suitable for deployment in web applications as a CAD system to assist surgeons in analyzing X-ray images. The model's ability to accurately predict "fracture", "metal", and "text" classes was high, with YOLOv9 achieving a 94.8% accuracy in "metal" class prediction. However, the performance in predicting bone anomalies and soft tissue classes was poorer, with YOLOv9 achieving only 11.3% and 28.2% accuracy, respectively. The paper also discusses the limitations of the current dataset and suggests future research directions to enhance model performance by improving the models with additional information on "bone anomaly" and "soft tissue" classes. The trained weight files are available on the authors' GitHub for further use. The study demonstrates that the YOLOv9 model can serve as a CAD system to assist radiologists and surgeons in interpreting X-ray images, particularly in pediatric wrist fracture detection.