An Extensive Review on Lung Cancer Diagnosis Using Machine Learning Techniques on Radiological Data: State-of-the-art and Perspectives

Abstract

Cancer of the lung is a killer illness that often results from a combination of genetic disposition and many clinical changes. The lung region of the body contains cancerous cells that develop atypically and threaten human life. To determine what might be helpful for its treatment, Lung cancer, also known as a Lung tumour, must be promptly and accurately discovered in its early stages. One of the most common malignancies and the main reason for cancer-related deaths globally is lung cancer. Its early detection offers a great chance to save lives. Therefore, automated lung cancer detection and classification from radiological data aided with machine learning and deep learning techniques have become essential because it provides a precise, trustworthy, and rapid investigation of the condition’s progression. This study summarises the current state of the art for automated lung cancer identification using machine learning from lung images, as stated in several recent publications. A thorough and extensive review was conducted after reading numerous research papers of latest 10 years from the compilation of this article, book chapters and articles Published inreputable national and worldwide databases, including PubMed, IEEE Xplore Digital Library, SpringerLink, Science Direct, Scopus and Google scholar. Several conference proceedings have also been included, provided they meet our quality review standards. The majority of the current algorithms enable an accuracy of almost 95% on openly available LIDC-IDRI, Lung Cancer data ‘data.world’ datasets. According to reports, deep Neural Networks, K-Nearest-Neighbors and Support Vector Machines are highly effective algorithms for classifying lung cancer from radiological images such as CT, MRI, and X-ray. For researchers and scholars in this discipline, this study not only highlighted obstacles but also highlighted and presented fresh research prospects. It has been generally recognised that while deep learning has mostly taken over the discipline of analysing medical images, classic machine learning methods like SVM and GMM performed excellently in classification. According to the articles that have been evaluated, the methodologies that are now in use still may need some improvements, which causes them to classify lung tumours with less accuracy. Furthermore, most algorithms today operate on a single type of radiological pictures like CT or MRI. This accuracy may be improved by changing several factors, like the features that need to be retrieved, improving noise removal, and employing hybrid segmentation and classification approaches like multi-level classifiers. Performance can be enhanced by combining the K-nearest-neighbours approach with other algorithms like support vector machines, pixel-level classifications, and extraction of regions of interest. This article will benefit the researchers working in this field to gain insight into how machine learning and deep learning techniques perform best when used with certain types of data collection, choice of features, numerous problems, and their suggested solutions in order to solve this challenging problem. At the conclusion of this review paper, limitations and potential areas for future study in the field of applying various machine learning approaches to the classification and diagnosis of lung cancer are also discussed.