Vision Transformer-based Classification for Lung and Colon Cancer using Histopathology Images

• Main Authors: Hasan, Munjur, Rahman, Md Saifur, Islam, Sabrina, Ahmed, Tanvir, Rifat, Nafiz, Ahsan, Mostofa, Gomes, Rahul, Chowdhury, Md
• Format: Conference Proceeding
• Language: English
• Subjects: Cancer; CNN; Deep learning; Histopathology; Hospitals; Lung; Lung cancer; Machine learning algorithms; ResNet; Transformers; VGG19; Vision Transformer
• ISSN: 1946-0759
• DOI: 10.1109/ICMLA58977.2023.00196

In this day and age, a considerable number of fatalities are caused by colon and lung cancer. Although their appearance simultaneously is rare, the likelihood of cancer cells spreading between these two organs in the absence of early diagnosis is relatively significant. The second highest cause of death worldwide is cancer. Colon and lung cancers are the most typical and lethal malignancies among the many different forms. Historically, physicians had to go through a lengthy and time-consuming process to evaluate histopathological images and identify cancer cases; however, current technology options make it possible to complete this process more quickly. This study classified the histopathological imagery LC25000 gathered at the Tampa, Florida-based James A. Haley Veterans Hospital of Lung and Colon Cancers using several Convolutional Neural Network (CNN) variants as well as the more recent vision transformer architecture. Based on the outcomes of this study, it has been observed that using the Vision Transformer provides exceptional performance outperforming deep learning CNN, VGG19, and Resnet 50 algorithms with 100% accuracy.