The timely identification of lung cancer is crucial for enhancing patient prognosis and overall treatment effectiveness. This work, proposed a predictive model that is U-Net and Res-U-Net for lung cancer by means of an IQ-OTH/NCCD dataset of CT images from a diverse patient population. In this work, three types of stages were used that is benign, malignant, and normal. The samples were taken for benign is 120, malignant is 561 and normal is 416. The proposed deep learning techniques, such as U-Net and Res-U-Net, have demonstrated notable performance, achieving accuracies of 87% and 97.6% respectively. This work has significant implications for improving early detection rates and ultimately advancing lung cancer diagnosis and treatment.