Early diagnosis of pulmonary disease is possible using Lung nodule detection in thoracic computed tomography (CT) scans. A fusion of U-Net and Dilated Convolution Networks (DCNs) has been proposed, for multi-scale lung nodule detection. U-Net architecture, utilized for semantic segmentation capabilities, is combined with DCNs to capture multi-scale features, to predict nodules of different sizes. The accuracy of the proposed work is better by the utilization of deep learning models. Proposed UNet-Dilated network based method is applied on thoracic CT datasets revealing better results, compared to traditional methods. Fusion of U-Net and DCNs proves to be an efficient solution for advancing lung nodule detection, contributing significantly to the field of medical imaging for improved patient care.