Deep Learning Model Based on Mamba U-Net for Ovarian Cancer Diagnosis

Ovarian cancer is one of the most lethal gynecological malignancies in women. Due to the absence of distinct symptoms in the early stages, it is often diagnosed late, which negatively impacts survival rates. Although current diagnostic methods rely on histopathological examinations, there remains a need for reliable biomarkers and computer-aided diagnostic systems for early detection. In this study, a deep learning model was developed and its performance was evaluated based on the Mamba U-Net architecture—an architecture found in the literature that combines the advantages of State Space Models (SSM) with the U-Net structure—for the purpose of ovarian cancer detection. After applying preprocessing and data augmentation techniques on a dataset consisting of histopathological whole-slide images, the model achieved an accuracy of 98.78%, a sensitivity of 97.83%, a specificity of 99.79%, a precision of 99.80%, and an F1 score of 98.80%. These results demonstrate that the model can make highly accurate and reliable predictions, highlighting its potential to contribute to the development of clinical decision support systems.