Deep Learning for Comparative Study of Ovarian Cancer Detection on Histopathological Images

Ovarian cancer is a serious threat to women's health. It comes from the ovaries, which are important for producing eggs and hormones, such as estrogen and progesterone. Ovarian cancer is usually a consequence of abnormal ovarian cell division, which leads to the creation of malignant tumors. The identification of ovarian cancer in its early stages can be difficult. Nonetheless, having an understanding of the detectable methods at hand is crucial in combating this illness. Deep learning is highly effective in detecting and distinguishing a variety of diseases through medical imaging. This study analyzes how effective deep learning architectures are at identifying cancerous and non-cancerous ovarian histopathological images. The study incorporates three distinguished deep learning models, including DenseNet-201, ResNet-50, and VGG-19. To evaluate models, important performance metrics, such as accuracy, sensitivity, specificity, and AUC, are used comprehensively. The observations show notable discrepancies in the models' abilities, with VGG-19 emerging as the leading structure, presenting exceptional accuracy and sensitivity. ResNet-50 demonstrates exceptional performance, particularly in specificity and AUC. Conversely, DenseNet-201, even though accomplishing a high level of accuracy, shows a need for improvement in specificity. The results show how deep learning can help diagnose cancerous and non-cancerous cells in the ovaries. They also suggest a model to use for classification.