DAE-DBN: An Effective Lung Cancer Detection Model Based on Hybrid Deep Learning Approaches

To effectively treat lung cancer and improve patient survival, automatic lung nodule identification is crucial. Computer-aided diagnosis (CAD) is one of the most efficient tools for quickly and reliably detecting lung cancer using an autonomous approach. Segmentation is an important and challenging task when it comes to diagnosing lung cancer from the LIDC-IDRI dataset. There are several factors that contribute to this, including shape irregularity, tissue inhomogeneity, and low contrast between the lung inner tissues and the surrounding tissues. Image segmentation techniques currently in use rely on various parameters, including image quality, tissue structure, and acquisition protocol, to accurately segment the different objects present in a lung's CT-scan image. Existing segmentation techniques rely on user input and expert evaluation to manually initialize the segmentation. However, this process is time-consuming, labor-intensive, prone to error, and not practical. Additionally, it negatively impacts the accuracy and efficiency of these approaches. This paper introduces the deep belief auto encoder model (DBA), which is a fully automated solution for detecting lung cancer. The model consists of two technologies: the Deep Auto encoder (DAE) and the Deep Belief Network (DBN). The DAE is responsible for extracting discriminative features from various objects found in LIDC-IDRI images. The features are subsequently utilized to construct the DBN network, which aims to identify the boundaries of distinct objects, thereby aiding in image segmentation. The experimental evaluation demonstrated that the proposed model achieved a high accuracy rate of 0.98%, which is significantly higher than the rates reported in previous studies.