Classification of Human Tissues from Histopathology Images Using Deep Learning Techniques

Classification of Nuclei from Histopathology images is an important step in both diagnosis and prognosis of disease. The proposed study classifies five different tissues that are annotated for their Nuclei from Histopathology images stained using Hematoxylin and Eosin. Various pre-trained deep learning-based models were deployed and the best results were obtained using the DenseNet121deep learning model. An accuracy of 90% and an Area Under Curve of 0.9843 was obtained in the classification of Histopatho images of five completely different tissues thereby showing the strength of deep learning in Whole slide image classification.