Ovarian cancer presents a diagnostic challenge because it does not show symptoms in the early stages and its diagnosis depends on the subjective interpretation of ultrasound images. The research team developed the EfficientOvaNet system, which uses deep learning to automatically classify ovarian tumors from ultrasound images. The model was trained on the MMOTU (Multi-Modality Ovarian Tumor Ultrasound) database and uses a two-branch EfficientNet-B3 architecture that combines details from specific regions of interest with the overall image context. The system incorporates advanced data processing techniques, including balancing uneven data distribution using weighted focal loss. Five-fold cross-validation was used to verify the reliability of the model. The EfficientOvaNet framework is equipped with explanatory tools such as Grad-CAM and uncertainty estimation that allow understanding of model decisions. According to the authors, this system could increase the accuracy of diagnosis, reduce subjectivity in evaluation and contribute to earlier intervention and improve the prognosis of patients with ovarian cancer.