A Study of Deep Learning Methods for Automatic Cancer Detection and Classification in Histopathological Whole-Slide Images

Histopathological examination is the diagnostic and research of tissue disorders. An integrated suite of histopathological samples has greatly aided doctors and researchers in the world of clinical research. Identifying cancerous tissues is critical for clinicians in providing appropriate oncology treatments. A whole slide image (WSI) is a digitized scanning of the tissues on the glass slide that allows the samples taken to be stored digitally on the computer system in the form of a digital picture. Cancer is among the top causes of death whereas lung and colon cancers are considered to be the deadliest with a very low survival rate. Lung cancer and colon cancer are considered to be devastating among all cancers. People affected by cancer mostly die. The survival rate is very less than approximately 5% in lung cancer and it is caused by various factors like smoking, drinking, etc. It not only affects the lungs but can spread into other parts of the body as well. India is witnessing more than one million cases of breast cancer per year. However, since histological images contain multiple tissue types and characteristics, classification is still challenging. WSI processing and storage have greatly aided professionals while encouraging researchers to develop more reliable and efficient fully automated analysis diagnosing models—ML, specifically DL-based models. DL models have outperformed in various disciplines, along with clinical applications and profound features in healthcare. A Fine-tune based framework is proposed and evaluated on two famous datasets KimiaPath24C and LC25000. The Kimia Path24 dataset was particularly created for the classification and retrieval of histopathology images and the LC25000 dataset for the classification of lung and colon cancer. The deep learning models like MobileNetV2, Xception, NasaNetLarge, EfficientNetV2L, and Self-build CNN (Convolutional Neural Network) model of 15 layers have been proposed on both datasets.