Towards Classification of Ovarian Cancer: A Vision Transformer Model

Ovarian cancer is a significant global health concern for women; early detection can save lives and improve patient survival. The variety in datasets and the growing complexities of deep learning methods have resulted in a continually increasing number of scenarios for these systems, complicating the correlation with various forms of ovarian cancer. In contrast to conventional convolutional neural networks (CNNs), which often inadequately grasp the overall context and struggle to capture comprehensive contextual information, transformer models demonstrate superior flexibility, enhanced contextual awareness, and an unparalleled capacity to analyze intricate global patterns. To address this challenge, two transformer learning systems have been developed for accurate ovarian cancer staging: Microsoft’s SWIN-Base Patch 4 and Google’s Vision Transformer vit-base-patch16-224. The dataset utilized for model training is sourced from online platforms. The vit-base-patch16-224 model outperformed the SWIN model, attaining 99% in precision, accuracy, recall, and F1-score. This study’s findings illustrate the revolutionary capacity of transformer models in medical imaging, laying the groundwork for progress in AI-enhanced healthcare.