Adversarial transformer network for classification of lung cancer disease from CT scan images

Abstract

Lung cancer is a dreadful disease that affects both men and women; an early prognosis is highly required to prolong human life. Recently, the computer-aided diagnosis (CAD) system has played a vital role in effectively diagnosing lung-related diseases. However, this system often fails to identify the type of cancers and is subjected to high error and computation complexity. Hence, this article introduces a novel deep learning (DL) technique to classify different types of lung cancers using chest CT images. The proposed study undertakes three major stages: pre-processing, feature selection and classification. The guided bilateral filtering (GWF) technique is initially introduced in the pre-processing stage to eliminate the noises from the raw CT images. Then, dimensionality reduction is performed using the weighted least absolute shrinkage and selection vector regression (We-LASSVR) technique to avoid over-fitting issues in the network model. In addition, the weighted mean of vector optimization (We-WMVO) technique is introduced to develop the performance of a proposed We-LASSVR technique. Finally, the transformer-aided generative adversarial network (T-GAN) technique is proposed to effectively classify the different types of lung cancers. Moreover, the dynamic levy flight chimp optimization (DyLF-CO) technique is proposed to tune the network model. The proposed model is employed in Python, and a chest CT image from the Kaggle website is collected and processed in this study. In the experimental section, the proposed method obtains an accuracy of 0.997, precision of 0.996, specificity of 0.998, RMSE of 0.104 and time complexity of 120 s.