In this study a deep learning method was developed that automatically segments and estimates endometrial volume from 3D ultrasound scans, addressing a key bottleneck in current IVF monitoring workflows. Trained on a large, diverse, multicenter dataset, the model achieves accuracy comparable to manual VOCAL measurements while reducing analysis time from several minutes to under two seconds. Its strong performance across different clinics and ultrasound systems demonstrates high robustness and generalizability. This technology has the potential to standardize endometrial volume assessment, support large-scale research on endometrial dynamics, and ultimately contribute to improved decision-making in assisted reproduction.