Automated segmentation of orthopedic implants in X-ray images is important for surgical planning, postoperative assessment, and implant analysis. We propose a fully automated detection-to-segmentation pipeline for knee implant X-rays in which YOLO-based detectors (YOLOv11 and YOLOv26) first localize implants and produce bounding boxes that are used to prompt the Segment Anything Model (SAM1) and enhanced SAM2.1 variants to generate segmentation masks. To our knowledge, this is among the first studies to evaluate YOLOv26 in a medical implant detection-to-segmentation framework, addressing a gap in the literature regarding the applicability of next-generation YOLO detectors for orthopedic X-ray analysis. We manually annotated high-quality ground truth masks for 310 images to enable quantitative evaluation and split the dataset into train/test sets using an 80/20 ratio. Performance was assessed using Dice, precision, recall, F1-score, and inference time. Using the same split ratio, training was repeated three times with different random seeds, and results are reported as the mean across runs. Within SAM1 and SAM2.1 settings, YOLOv11 consistently outperformed YOLOv26 while achieving comparable inference time, indicating that detector choice substantially influences end-to-end segmentation quality, suggesting that localization improves box-prompted segmentation even when the segmentation model is held constant.