DeepDenseVit: A Hybrid Deep Learning Approach for Ovarian Cancer Classification

Ovarian cancer, one of the most threatening gynecological cancers, but often it goes undetected due to non-specific symptoms. Early identification is crucial for improving survival outcomes. For identifying the disease in its early stages, the traditional diagnostic methods such as CA-125 blood tests and pelvic exams frequently fall short. The aim of this research is to fill the gap by utilizing advanced machine learning techniques to enhance ovarian cancer identification. By using a dataset of 987 histopathological images, our study proposed DeepDenseVit comparing mostly used pretrained model such as DenseNet-121, EfficientNet B0, VGG 19 models. Our proposed DeepDenseVit model combined DenseNet-121 and Vision Transformer (ViT) with custom layers for identifying ovarian cancer. DeepDenseVit demonstrated exceptional performance, achieving a test accuracy of 95%, precision of 94.99%, recall 94.69% and F1 score of 94.98%. The result shows the effectiveness of the proposed model for identifying ovarian cancer.