Background/Objectives: This study aims to create a strong binary classifier and evaluate the performance of pre-trained convolutional neural networks (CNNs) to effectively distinguish between benign and malignant ovarian tumors from still ultrasound images. Methods: The dataset consisted of 3510 ultrasound images from 585 women with ovarian tumors, 390 benign and 195 malignant, that were classified by experts and verified by histopathology. A 20% to80% split for training and validation was applied within a k-fold cross-validation framework, ensuring comprehensive utilization of the dataset. The final classifier was an aggregate of three pre-trained CNNs (VGG16, ResNet50, and InceptionNet), with experimentation focusing on the aggregation weights and decision threshold probability for the classification of each mass. Results: The aggregate model outperformed all individual models, achieving an average sensitivity of 96.5% and specificity of 88.1% compared to the subjective assessment’s (SA) 95.9% sensitivity and 93.9% specificity. All the above results were calculated at a decision threshold probability of 0.2. Notably, misclassifications made by the model were similar to those made by SA. Conclusions: CNNs and AI-assisted image analysis can enhance the diagnosis and aid ultrasonographers with less experience by minimizing errors. Further research is needed to fine-tune CNNs and validate their performance in diverse clinical settings, potentially leading to even higher sensitivity and overall accuracy.

Background/Objectives: Accurate preoperative characterization of ovarian masses is essential for appropriate clinical management, particularly for borderline ovarian tumors (BOTs), which are less common and often difficult to distinguish from benign or malignant lesions on ultrasound. Although expert subjective ultrasound assessment achieves high diagnostic accuracy, limited availability of highly trained sonologists restricts its widespread application. Artificial intelligence-based approaches offer a potential solution; however, the low prevalence of BOTs restricts the development of robust deep learning models due to severe class imbalance. This study aimed to develop a Convolutional Neural Network (CNN)-based classifier enhanced with Generative Adversarial Networks (GANs) to improve the discrimination of ovarian masses as benign, malignant, or BOT using ultrasound images. Methods: A total of 3816 ultrasound images from 636 ovarian masses were retrospectively analyzed, including 390 benign lesions, 202 malignant tumors, and 44 BOTs. To address class imbalance, a Deep Convolutional GAN (DCGAN) was used to generate 2000 synthetic BOT images for data augmentation. A three-class ensemble CNN model integrating VGG16, ResNet50, and InceptionNetV3 architectures was developed. Performance was assessed on an independent test set and compared with a baseline model trained without DCGAN augmentation. Results: The incorporation of DCGAN-generated BOT images significantly enhanced classification performance. The BOT F1-score increased from 68.4% to 86.5%, while overall accuracy improved from 84.7% to 91.5%. For BOT identification, the final model achieved a sensitivity of 88.2% and specificity of 85.1%. Class-specific AUCs were 0.96 for benign lesions, 0.94 for malignant tumors, and 0.91 for BOTs. Conclusions: DCGAN-based augmentation effectively expands limited ultrasound datasets and improves CNN performance, particularly for BOT detection. This approach demonstrates potential as a decision support tool for preoperative assessment of ovarian masses.