Uterine diseases, including adenomyosis, fibroids, endometriosis, and cancers, are among the most prevalent gynecological conditions, significantly impacting women’s health worldwide. Accurate diagnosis depends on imaging modalities such as Magnetic Resonance Imaging and Ultrasound. However, traditional manual segmentation methods are time-consuming and prone to variability, limiting their clinical utility. This study reviews recent advancements in Artificial Intelligence (AI)-driven image segmentation for uterine disease diagnosis, focusing on deep learning models such as U-Net, nnU-Net, and Vision Transformers. Findings from ten key studies are synthesized, examining methodologies, datasets, preprocessing techniques, and evaluation metrics. State-of-the-art models achieved Dice similarity coefficients between 0.82 and 0.92, demonstrating superior performance over traditional approaches.