A systematic review and meta-analysis on performance of intelligent systems in lung cancer: Where are we?

Abstract

Cancer is one of the human’s life-threatening diseases that does not merely pertain to one organ. Despite the varieties of cancers, lung cancer, with its different growth and spreading mechanisms, can affect the normal cells and disrupt the cell signaling procedure that alters the cell division function. In this systematic review and meta-analysis, the well-known databases are searched based on a Boolean query exclusively for lung cancer and the corresponding artificial intelligent systems. By systematically searching the PubMed and Scopus databases, English-language articles published up to 13 July 2017 were extracted that identify the cancerous and normal cell images using different types of predictive models. Then, the search results will be selected for the pertinent articles encompassing the required information (i.e., inclusion criterion) such as total sample size, true positive, true negative, false positive, and false negative values. The studies without enough information were omitted from further analysis. Considering the lung cancer diagnosis and conducting the meta-analysis on the articles, the results for the improvement trends in the amount of success in the performance of the artificially intelligent systems have been reported. Eventually, two publication bias tests have shown that the possibility of publication bias exists. And, the trends on diagnostic odds ratio and AUC values were immeasurably high, respectively, while those of sensitivity and specificity were moderate.