Deep Learning-Driven Precision Diagnosis of Ovarian Cancer in Histopathology Images
Janani V G, S. Vasuki, T. Haley Pearl Caxmi, S. Pritthiga · 2025
Recent advancements in medical image analysis have improved ovarian cyst detection using histopathological images. Building on this, a hybrid deep learning model combining VGG19 and Xception is introduced for more accurate ovarian cancer detection. This integration leverages VGG19's structured hierarchical feature extraction and Xception's efficient depthwise separable convolutions, enhancing classification performance. It captures fine-grained patterns in high-resolution images, while Xception improves computational efficiency by reducing redundant parameters. This hybrid approach ensures both low-level textures and high-level semantic patterns are effectively utilized for precise detection. Data augmentation increases dataset diversity, addressing the challenge of limited labeled histopathology images and improving generalizability across different medical facilities. The fusion of VGG19 and Xception allows for comprehensive feature extraction, leading to superior classification accuracy and sensitivity in detecting ovarian cancer compared to a single model. Our proposed algorithm simulation results show the hybrid approach outperforms standalone VGG19 and VGG16 in distinguishing between benign and malignant cysts, ensuring more reliable diagnostics for pathologists. This work highlights the impact of deep learning in improving ovarian cancer detection, demonstrating how hybrid models enhance medical image analysis. The combination of VGG19 and Xception optimizes detection accuracy and contributes to early diagnosis, ultimately improving patient outcomes.