Persistent Homology (PH) has gained popularity in topological data analysis. Building on this trend, we introduce PH-SAM, a novel hybrid approach for segmenting X-ray images with targeted implants in medical imaging. Our method leverages PH and the Segment Anything Model (SAM) which employs the Convolutional Neural Network (CNN) and the Vision Transformer (ViT), all powerful tools, to efficiently compute image feature profiles. These profiles capture the distinct characteristics of regions of interest (ROI), such as knee implants captured in X-ray radiographs, enabling precise discrimination between implants and normal regions. Our primary goal is to achieve automated, and higher accuracy in knee X-ray implant segmentation. To accomplish this, we propose an innovative technique that combines PH with SAM, harnessing the strengths of both approaches. By integrating PH, the SAM component no longer requires manual interaction. Instead, PH automates the identification of the region of interest, allowing the SAM prompt encoder to work without user intervention. PH-SAM automatically and effectively captures both topological and morphological features in medical scan images. The PH-SAM test result Dice Sore for an individually processed image was ≅ 0.98 and the average of 35 tested images Dice was ≅ 0.80. The performance of PH-SAM holds promises for various image segmentation tasks, highlighting its effectiveness and potential for broader applications in medical imaging and other areas requiring autonomous and efficient segmentation such as Airport Body Scanning (ABS) machines.