Automatic Lymphocyte Detection on Breast Cancer Histological Images Using Deep Learning

Breast cancer is the most common and fatal form of cancer among women. Therefore, it becomes essential to diagnose it quickly for appropriate treatments. The lymphocyte detection in the histological images has become increasingly important in therapeutic disease diagnosis and monitoring. We analyze the set of histological images of breast tumours taken from the BreCaHad dataset and we concentrate on the detection of lymphocytes. To this aim, we design a process consisting of two steps: (i) a segmentation step, obtaining a mask isolating the cells in the histological images by a deep learning model based on a convolutional neural network; (ii) a classification step, identifying the presence of lymphocytes by a binary classifier trained on the cells isolated at the previous step. The best classification performance was reached by the random forest model (Fl-score value of 93.13% and an accuracy of 93.20%).