Sooty-LuCaNet: Sooty tern optimization based deep learning network for lung cancer detection

Abstract

Lung cancer is the prevalent malignancy afflicting both men and women, mostly affects the chain smokers. The lung CT images are examined to identifying the abnormalities, but diagnosing lung cancer with CT images is time-consuming and difficult task. In this work, a novel Sooty-LuCaNet has been proposed in which the best features are selected using sooty tern optimization to reduces computational complexity of neural network. Initially, the denoised CT images are segmented using Grabcut technique to separate the lung nodules by eliminating the background distortions. The deep learning based Shufflenet is used to extract the structural features from the segmented nodule and the textural features from the enhanced images. Afterwards, the sooty tern optimization (STO) algorithm is applied to select the most relevant features from the extracted features from the ShuffleNet. Finally, the classification process is carried out to differentiate the normal, small cell lung cancer (SCLC) and non-small cell lung cancer (NSCLC) from the CT images. The experimental findings show the robustness of the proposed Sooty-LuCaNet based on the specific metrics namely sensitivity, accuracy, specificity, recall, precision and F1 score. An average classification accuracy of 99.16% is achieved for detection and classification of lung cancer.