Abstarct: Abstract Lung cancer is one of the diseases that is currently a leading cause of death for many people worldwide. According to statistics published in 2016, it is the deadliest type of cancer among all existing cancers. Pulmonary nodules are the primary indicators of lung cancer, and timely identification of these nodules significantly impacts the treatment outcomes. In recent decades, the use of artificial intelligence methods and image processing algorithms has become a key approach for detecting and diagnosing pulmonary nodules. Identifying a nodule in medical images involves determining its location within the image, a process known as segmentation. Segmentation refers to the operation of separating an image into its constituent components. Classical image processing algorithms, such as edge-baxsed and region-baxsed segmentation methods, were widely used before the emergence of deep learning techniques. These methods were fundamentally baxsed on thresholding and were suitable in terms of clarity, simplicity, and execution speed; however, they required manual adjustments and feature extraction by the user. With the advent of deep learning and the increase in available datasets, the application of deep convolutional networks for medical image detection has garnered significant attention. The aim of this research is to segment pulmonary nodules using deep learning techniques. For this purpose, we utilized CT scan images of the lungs through a hierarchical process consisting of three main stages to identify the nodules. Initially, in the preprocessing stage, histogram equalization of the images was performed. The second stage involved using a deep network called BCD-UNET to extract the lung region from the CT images. Finally, in the third stage, the location of the nodule within the lung region was identified using an Attention-Unet network that received the output from BCD-UNET. The results obtained from the LUNA16 dataset indicate a high accuracy of the algorithm at each stage of the proposed method. Specifically, the BCD-UNET network achieved DICE and IOU scores of 97.75% and 95.79%, respectively, while the Attention-Unet network attained DICE and IOU scores of 91.73% and 90.39% in nodule extraction. Additionally, the sensitivity metric for accurately determining the identified nodules was found to be 92.31%.