Inter-observer variability, lack of standardization, the necessity to document incidental findings, and the clinical demand for rapid diagnostic support limit the efficiency and reliability of pelvic MRI, emphasizing the need for automated analysis of uterine MRI scans. This work introduces a real-time, deep learning-based tool designed to automatically generate structured analysis reports directly from sagittal T2-weighted pelvic MR images. Utilizing real-time scanner interfacing and state-of-the-art 3D nnU-Net architecture with a Residual Encoder preset trained on a publicly available dataset, the proposed system accurately segments the uterine wall, uterine cavity, uterine fibroids, and Nabothian cysts. Post-processing of the predicted segmentation enables the generation of comprehensive structured HTML reports that include precise uterine volumetric measurements and detailed lesion assessments for fibroids and Nabothian cysts. The performance of the tool was validated on two independent datasets from different clinical sites, varying in magnetic field strength, and scanner vendor. Its real-time inference capability was also confirmed. The segmentation model showed reliable performance uterine wall, uterine cavity and uterine myoma (mean dice coefficient of 0.82, 0.77 and 0.78 respectively) and a mean dice coefficient of 0.43 for Nabothian cysts. Reports were consistently generated in real-time within an average time of 60 s. By providing immediate, standardized, and reproducible analyses, the developed tool is well-positioned for seamless integration into clinical radiological workflows.