In the world, one of the most critical diseases is lung cancer that leads to death of almost all affected human beings due to uncontrolled growth in the cell. These abnormal cells grow rapidly and divide to form tumor in the lungs. For lung cancer detection, the CAD system divided in four parts in CT images, such as noise removing process, segmentation of lung, detection of lung nodule and classification. The Visual information of similar nodules helps radiologists to detect the disease. This paper contains the Content-Based Image Retrieval (CBIR) system which is used for nodule retrieval in lung CT images. The CBIR system of pulmonary nodules retrieval system consists of pre-processing, feature extraction, feature selection, retrieval and classification. Our proposed method used manual cropping for segmentation phase to extract Region of Interest (ROI). The Gray-Level Co-Occurrence Matrix (GLCM) is used to extract features and the extraction of feature is done by Fast Discrete Curvelet Transform (FDCT). To select the perfect features, enhanced moth flame optimization algorithm is used and other best features are filtered by Deep Neural Networks (DNN's). We used the Euclidean distance to retrieve similar lung nodules from ROI database. The proposed method has been tested on the LISS database. Finally, we have achieved the content based image retrieval from ROI image database.