Aprendizaje automático para el diagnóstico de células cancerosas en imágenes citológicas de líquido pleural: Una revisión sistemática de la literatura

Machine learning is used in medicine to diagnose diseases quickly and accurately, the results of which support the physician in making correct decisions. Pleural effusion, a common disease in which 50% of patients are diagnosed with cancer. The objective was to describe machine learning techniques that are used for the diagnosis of cancer cells in cytological images of pleural fluid. For the systematic review, the PICO strategy and the PRISMA methodology were used. For the research questions, selection criteria were established, identifying 142 articles, selecting 18 articles after filtering. The techniques used were U_Net with 8 articles, Transfer Learning with 4 articles, Support vector machine with 3 articles, CNN with 3 articles, ANN with 3 articles, X-Boost with one article, K-Means with one article and other ML techniques with 4 articles. Regarding the data set, the most used were cytological images in 10 investigations, CT images in 4 investigations, X-ray images in 3 investigations and one investigation with 1 ultrasound. This literature review will support future research to apply the models and techniques, since there are not many studies on the prediction of cancer cells in pleural fluid.