A Robust Deep Learning Framework for Ovarian Cancer Subtype Classification
Sihan Lv, Gaofeng Zhang · 2024
Accurate identification of ovarian cancer subtypes from histopathological images is crucial for guiding treatment, but remains challenging. We present a deep learning approach to classify ovarian cancer subtypes from whole slide images (WSIs) and tissue microarrays (TMAs). Our EfficientNetV2 based model achieved 93.08% balanced accuracy on WSIs, outperforming prior methods. To enable TMA inference, we developed an algorithm to generate a TMA-like dataset from WSIs, allowing our model to achieve 80% accuracy on a true TMA test set. This highlights the robustness and clinical potential of our framework for aiding ovarian cancer diagnosis and treatment selection.