Artificial Intelligence applied to asthma biomedical research: a systematic review

Abstract

Introduction: A major application of Artificial Intelligence (AI) is to uncover relevant information from big data. This technology could play major roles in medicine, such as identification of new targets, discovery of new molecules, diagnostics, therapy selection, risk prediction and stratifying disease. Aims and Objective: To provide a review of existing algorithms for the application of AI in research and medical management of asthma. Methods: We performed a systematic review of English scientific articles, using the PubMed database, until Dec. 2018. Search terms included AI, machine learning, deep learning in single combination with asthma term. We included papers focused on human asthma, based on machine learning algorithms. Results: We selected 136 papers on 253 found after excluding duplicated and papers which did not meet inclusion criteria. 52 (40%) regarded the application of AI in asthma pathway analysis, phenotype and biomarker identification, 77 (56%) involved AI in asthma diagnosis, early prediction of exacerbations and predicting control, 7 (5%) are related to AI as support to the management and personalization of the treatment. Conclusions: Standard validation method of these technologies has not been established and data used in each work originate from different sources. Hence it is impossible to perform a direct outcome comparison of selected articles for each application. Evidences of AI confirmed proof of concept, but in order to transfer AI to clinical practice a systematic evaluation of properties, effects, and impacts of health technology is needed.