Computer Aided System for Nuclei Localization in Histopathological Images Using CNN

Today, the health care industry is extensively using computer aided diagnostic expert system. The expert system for the diagnosis of breast cancer using the histopathological image is the need of time. Analysis of histopathological images is challenging due to its complex architecture with irregularly shaped nuclei. Convolutional neural network (CNN) is a promising technology emerging in recent years. We have designed a computer based expert system to identify nuclei in histopathological images. The system is developed using python programming language. We have used the BreaKHis breast cancer dataset for experimentation and Kaggle dataset for convolution masks generation. Nucleases are localized using custom design Keras and U-Net Hybrid CNN (KUH-CNN) model. The systems can be used by histopathologists for the diagnosis of malignancy in the tissue. The system can also aid the researchers who can implement a machine learning algorithm on the nucleases detected images for further analysis.