A Review of Deep Learning Techniques for the Detection of Lung Cancer on Medical Images: Current Advancements and Challenges

Abstract

Lung cancer is a prevalent global ailment characterized by the uncontrolled proliferation of aberrant cells within the lungs. Frequently, it happens within the airway cells, particularly in the bronchi, which serve as the main conduits for air to reach the lungs. While it is not possible to completely avoid lung cancer, it is indeed possible to mitigate the risk associated with it. The occurrence of cancer has resulted in an increase in death rates for both males and females. Patients have a far better chance of surviving if lung cancer is detected early. Computed Tomography (CT) is a commonly employed imaging modality for distinguishing between malignant and benign lung nodules. However, the cost of a CT scan was prohibitively high, making it unaffordable for many people in remote areas. In addition, the process of analysing those scans require a significant amount of time and a team of highly qualified radiologists. Lung cancer detection and classification using CT images has seen a proliferation of Deep Learning (DL) frameworks in the last several decades.  These algorithms have the capability to identify dubious nodules or lesions at an early stage, which could potentially result in earlier identification and enhanced patient outcomes. Furthermore, the insights derived from those frameworks might assist clinicians in making informed decisions and achieving early diagnosis, therefore mitigating the likelihood of adverse patient outcomes. This research provides an extensive analysis of various DL frameworks that have been built for the purpose of identifying and classifying lung cancer based on CT images. Firstly, a brief study is conducted on various lung cancer categorization systems developed by multiple researchers, which are based on DL algorithms. Afterwards, a thorough evaluation is carried out to determine the shortcomings of current algorithms and provide a fresh strategy for precise lung cancer classification, with the goal of reducing global mortality rates.