This report surveys ensemble deep learning approaches for automated osteoarthritis (OA) grading in knee X-ray images. OA grading is currently performed manually using the Kellgren-Lawrence (KL) scale but is limited by subjectivity. Ensemble deep learning can improve accuracy by combining multiple models. The key ensemble methods for OA grading are analyzed, including bagging, boosting, and stacking. Studies applying these methods using datasets like the Osteoarthritis Initiative are reviewed in detail. The report discusses challenges like lack of generalizability and the need for interpretability. Potential applications in clinical practice including early detection and reducing diagnostic variability are explored. However, issues around data privacy, model integration into workflows and lack of trust also require attention. Overall, ensemble deep learning holds promise for OA grading but addressing these challenges is important for successful adoption. The ensemble deep learning progress along the way it is expected that collaborative learning techniques will become more significant in the detection and treatment of osteoporosis.