Lung cancer represents one of the most severe forms of tumors in individuals, necessitating an intricate and time-consuming research procedure for identification and classification. The challenge is compounded by the diverse nature of lung nodules and their visual resemblance to adjacent areas. Traditional machine learning approaches either handle these factors in isolation or rely heavily on human assimilation, potentially missing intricate relationships among features. By leveraging the layered structures of deep learning methods, a novel architecture introduces a proficient practice for the segmentation and classification of pulmonary nodules in Computerized Tomography (CT) images. The methodology begins with comprehensive data pre-processing methods to ensure data readiness. Subsequently, a T-net model-based module is employed to segment lung nodules, while a Centernet-based approach procures intensity and textural traits from the segmented images. Following this, a Nasnet-based classification module labels nodules as cancerous or non-cancerous based on the collected attributes. The performance of the method is then evaluated on the Lung Image Database Consortium dataset using various metrics. For the Segmentation task, it achieved parameters (Sensitivity: 97.40, Positive Predictive Value: 96.02, and Dice similarity coefficient: 97.63). For the Cancer Classification task, it achieved parameters (F1-score: 99.21, Precision: 99.22, Recall: 99.20, and Accuracy: 99.24). This innovative approach not only improves the efficiency and accuracy of lung nodule Segmentation but also lays the groundwork for enhanced diagnostic capabilities in lung cancer detection and management.