This study developed a patient-specific cardiac motion-based virtual reality cardiovascular intervention simulator for training purposes. Personalized 3D cardiac models were generated from medical images using AI-based nnU-Net segmentation, and ECG-synchronized motion was integrated to reproduce physiological cardiac cycles through P-QRS-T waveform analysis. In particular, this study went beyond simple simulation by designing and implementing serious game elements including difficulty adjustment by user skill level, real-time haptic feedback, and a quantitative scoring system to maximize educational effectiveness. Real-time stent insertion was implemented at performance exceeding 60fps through the Extended Position-Based Dynamics (XPBD) algorithm. Experimental results showed that the segmentation model achieved high accuracy with an average Dice Similarity Coefficient (DSC) above 0.90, and the dynamic model demonstrated biomechanical behavior similar to clinical data, showing 12.3% coronary artery diameter change and 3.2mm positional displacement during the cardiac cycle.