Mitosis Detection in Breast Cancer with Deep Learning: A New Approach

In this work, we propose a new approach for spot-
ting mitoses in breast cancer histology images. This new approach
involves integrating two publicly accessible datasets following a
normalization procedure of color. The mitotic samples are then
enhanced by preserving the context to address class imbalance.
After this, the candidate mitotic cells are classified into the target
classes using a ResNet classifier. Through this method, we were
able to accurately identify mitosis in the combined dataset while
attempting to identify it in the images. We demonstrate that our
method outperforms all current methods by comparing it to state-
of-the-art methods using a public dataset. Our results indicate
that the proposed technique can be used to automatically identify
mitotic cells in the images of histopathology of breast cancer.
Index Terms—histopathology, ResNet-50, mitosis detection