Introduction: A prevalent malignant tumor illness with significant mortality and clinical impairment rates is lung cancer. Currently, manual pathology section analysis is the primary method used to detect lung cancer, although this method is inefficient and vulnerable to error due to its subjective character. There are several imaging techniques to diagnose lung cancer, such as computed tomography (CT) scan, Magnetic resonance imaging (MRI) scan, and chest X-ray. Considering the high rate of false positives and negatives, human error in the interpretation of images, low speed of analysis of results in emergencies, and low accuracy in these methods, it seems that we need auxiliary methods to solve these defects. With the ongoing development of technology, artificial intelligence (AI) has gradually been included in imaging diagnosis, and it has the potential to improve the effectiveness of lung cancer screening. Several studies in this field have been conducted on several imaging methods, which will be comprehensively reviewed. This review deals with the question of whether artificial intelligence has sufficient sensitivity and specificity to outperform human experts in diagnosis. Also, this study aims to express the advantages and problems of artificial intelligence as a tool for better analysis of imaging results.
Methods: This article is written as a review. A specific search strategy was determined based on keywords and their synonyms. Then the articles were extracted by searching Google Scholar and PubMed databases. Keywords included lung cancer and artificial intelligence, and only primary studies including interventional and observational were studied. The obtained articles were filtered based on specific inclusion and exclusion criteria and qualitatively reviewed. Finally, the results of the selected articles were reported by mentioning the methodology and main findings of the study.
Results: Based on the results, the diagnostic system with the help of artificial intelligence for imaging techniques has a significant diagnostic accuracy (with p value less than 0.05) for the diagnosis of lung cancer, which has a significant value for the diagnosis of lung cancer and more possibility to realize the development application in the field of clinical diagnosis. Therefore, AI with intelligent learning algorithms and high accuracy can detect imaging findings that are usually missed. However, there is a margin of error and the clinical utility of AI has yet to be fully proven.
Conclusion: AI has surfaced as a highly promising instrument to aid radiologists in the examination of thoracic images for the detection of lung cancer. The advantageous features of AI comprise the proficiency to expeditiously process copious amounts of data, detect subtle anomalies that may elude human observers, and provide quantitative measurements for precise diagnosis. Furthermore, AI algorithms exhibit the capability to acquire knowledge from vast amounts of data and progressively elevate their efficiency over a period of time. Several matters necessitate resolution to effectively employ artificial intelligence in the identification of lung cancer. To train AI models, for instance, one needs high-quality annotated datasets. Other requirements include ensuring the robustness and generalizability of algorithms across various populations and imaging modalities, as well as addressing moral and legal issues and gaining acceptability.