O uso do diagnóstico assistido por computador na pesquisa da malária: uma avaliação entre o padrão ouro de diagnóstico da malária e o sistema de diagnóstico automatizado por meio da inteligência artificial

Abstract

Background: Malaria is one of the most serious parasitic diseases that affect a large part of the world's population. Transmitted through the bite of a female mosquito of the genus Anopheles, which deposits the parasite into the bloodstream, the disease is only detected through a precise examination analysis, and since microscopy, even though it is the gold standard for diagnosis, has interferences that can compromise the analysis. Therefore, the automated method that uses computer-aided diagnosis technique (CAD) through the use of machine learning models to aid microscopy, can bring more reliable and consistent results. Objective: In this research, a comparative review between the use of the CAD system and the gold standard tests for diagnosing malaria. Method: In this way, a survey of existing studies on the use of the CAD system in the diagnosis of malaria was carried out on the main search platforms: MEDLINE/PUBMED, ScienceDirect, IEEE, Cochrane Library, Portal de Periódicos da CAPES, LILACS, EMBASE and Google academic; The search used previously defined expressions in English. The articles located were subjected to preliminary cataloging of publications, the application of a selection filter for relevant publications (inclusion and exclusion criteria) and a quality filter. Analyzes of statistical variables, such as sensitivity and specificity, were carried out and compared with the results of gold standard tests, thus observing a significant statistical difference between the groups. Conclusion: According to the research carried out, they showed satisfactory results for the implementation of the CAD system in the diagnosis of diseases. The aggregation of deep learning methodologies with CNNs has enhanced and improved results in the identification of malaria parasites compared to microscopy techniques.