LDDC-Net: Deep Learning Convolutional Neural Network-based lung disease detection and classification

Abstract

Lung cancer is the crucial disease, which causing to millions of deaths around the globe.  Therefore, the early detection and classification of lung cancers can save millions of lives. However, the conventional methods were failed to result the better classification performance. Thus, this is implemented deep learning convolutional neural network (DLCNN) model for lung disease detection and classification operations (LDDC-Net). Initially, preprocessing of Computed Tomography (CT) based lung images were performed using modified non-local trilateral filter (NLTF). Then, segmentation of lung cancer is performed using hybrid fuzzy morphological (HFM) operations, which effectively localizes the region of interest (ROI) of cancer. Further, laplacian pyramid decomposition (LPD) process applied on segmented image to extract the deep seismic features. Further, grasshopper optimization algorithm (GOA) based evolutionary model is used to select the best features. Finally, DLCNN is model performed training, testing operations using extracted features and classifies the benign, malignant lung cancers. The simulation results shows that the proposed LDDC-Net resulted in superior segmentation, classification performance as compared to conventional methods