Leveraging Medical Foundation Model Features in Graph Neural Network-Based Retrieval of Breast Histopathology Images
Breast cancer is a prevalent health issue, and early detection is vital. To support pathologists, we propose a novel image retrieval framework leveraging foundation model features and an attention-based adversarially regularized variational graph autoencoder. Our method achieve state-of-the-art results on the BreakHis and BACH datasets. This framework has the potential to enhance diagnostic workflows in clinical settings.