Vision Transformer Based Effective Model for Early Detection and Classification of Lung Cancer

Download PDF
Open Webpage

Arvind Kumar, Ravishankar Mehta, B. Ramachandra Reddy, Koushlendra Kumar Singh

Published: 2024, Last Modified: 06 Nov 2025SN Comput. Sci. 2024EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: This study explores the worldwide effects of lung cancer and its early detection and diagnosis. Artificial intelligence (AI)-based models are quite popular among researchers in this field for early detection of lung cancer. Histopathology images are one of the popular means for diagnosis and detection of lung cancer. The present work encompasses the Vision Transformer (ViT) based model to classify lung cancer based on histopathological images. The proposed ViT-based model has been shown to have a promising impact in deciphering complex spatial relationships within image data. The proposed model has been validated by a publically available database, namely the LC25000 dataset, which contains lung and colon cancer histopathology images with variations in tissue types. The primary intuition behind using this ViT is to freeze the pre-trained ViT layers as a feature extractor and perform classification tasks by adding a new classification head. The model has been tested with various patch sizes during the model training. The proposed method achieved the best accuracy of 98.84% when the patch size was set to 16 × 16. Furthermore, the efficiency of the proposed work has been tabulated and compared work with existing work.