Detecting and Classifying Atypical Teratoid/Rhabdoid Tumors using YOLO-v11 and VGG-19

This research presents a snapshot deep learning method to identify atypical teratoid/rhabdoid tumors (AT/RT) in pediatric patients using MRI data. The procedure considers VGG-19 for classification and YOLOv11 for real-time tumor location. VGG-19 gave around 92% accuracy in differentiating AT / RT from healthy tissue, whereas YOLOv11 realtime detection accuracy was 92.5%, suggesting its relevance for time-sensitive clinical situations. Scaling, normalization, and data augmentation helped make the model more resilient. Transfer learning using pre-trained weights from ImageNet (VGG-19) and COCO (YOLO-v11) was utilized for more efficient training. Models were then trained for about 40 epochs using tuned hyperparameters. Precision, recall, and F1-score measures validated the performance of the model. This research demonstrated that VGG-19 is good for diagnostic classification, and YOLO-v11 works well for making real-time cancer diagnoses. This method allowed early and reliable AT/RT diagnosis in pediatric patients-to the advantage of treatment outcome. Future work will look at other rare pediatric brain tumors and multi-modal imaging.