An persistent autoimmune disease is rheumatoid arthritis (RA) that significantly impacts joint health that, if not identified and treated right away, may result in significant morbidity. Early detection through medical imaging, particularly X-ray, is vital for timely intervention and better patient outcomes. Since there are four stages of RA, it is very hard to fix the detection of it in the earlier stages. A promising approach for automating the detection and classification of medical pictures is the use of convolutional neural networks (CNNs), but optimizing these networks for complex, high-dimensional data remains challenging. This study introduces a novel approach to predicting RA from X-ray images using a CNN architecture; the hyperparameter of the CNN has been optimized by the Grey Wolf Optimization Algorithm (GWO). Inspired by the social hierarchy and hunting strategies of grey wolves, GWO efficiently addresses complex optimization tasks. The methodology includes preprocessing X-ray images to emphasize features pertinent to RA diagnosis and training the CNN using GWO optimization. In order to ensure a comprehensive assessment of the suggested approach, the dataset includes X-ray pictures from both RA patients and healthy controls. The efficacy of the CNN-GWO model is evaluated using performance parameters like accuracy, sensitivity, specificity, and AUC-ROC. Comparative analyses highlight its superiority over traditional CNN approaches, marking a significant advancement in automated RA diagnosis and offer the way for improved diagnostic tools in medical community.