Rheumatoid arthritis (RA) is a chronic autoimmune disorder with persistent joint inflammation. The incidence of RA increases with age. Medical imaging plays a critical role in the diagnosis and monitoring of RA by facilitating the detection and evaluation of various pathological changes associated with the disease. Over the past two decades, computer-aided diagnostic (CAD) systems based on artificial intelligence (AI) have been extensively studied with the aim of integrating them into the radiological analysis of RA, particularly in quantifying the progression of joint space narrowing (JSN) and bone erosion on conventional radiographs, which have significantly alleviated the workload of rheumatologists. By employing algorithms for classification, detection, and image registration, CAD systems enable more accurate and efficient analysis of radiographic features indicative of RA, enhancing the precision and speed of diagnosis and supporting earlier and more targeted therapeutic interventions. Despite these advancements, several challenges persist in developing CAD systems for RA. One major obstacle is the complexity of accurately annotating joint space, which requires precise localization of bone edges with sub-pixel accuracy in radiographic images. This task is far more challenging than standard mask annotations, often resulting in inconsistencies or inaccuracies in manual labeling. Moreover, the scarcity of high-quality annotated data remains a significant bottleneck in advancing CAD tools in this domain. Another critical issue is the imbalanced distribution of joint space width (JSW) data in existing datasets, where samples predominantly represent early-stage RA while advanced-stage cases are underrepresented. This imbalance impedes the training of AI models, limiting their ability to learn and predict JSW variations accurately across the full spectrum of RA progression. Generating synthetic data in joint images offers a promising solution to address these challenges. Such synthetic data can supplement existing datasets, enhancing the performance and robustness of AI models in RA-related joint space analysis and ultimately contributing to more reliable and effective diagnostic tools in clinical practice.