Nuclei Segmentation in Hematoxylin and Eosin (H&E)-Stained Histopathological Images Using a Deep Neural Network

Breast cancer, being the second most frequently seen cancer globally, is the most common cancer among women. Early and accurate diagnosis is dependent on the correct identification of nuclei in histopathological images, which is a very tedious task when it is done manually. The proposed method is a deep learning model using U-Net architecture supplied with various pre- and post-processing techniques. The model was trained and tested on a unique dataset. Our model performed the nuclei segmentation task with a Dice score of 0.84, a true negative rate (TNR) of 0.95, a true positive rate (TPR) of 0.84, and an accuracy rate of 0.93. To enhance the performance of our model, our planned future work includes increasing the number of histopathological images in our dataset.