Fuzzy-ART neural networks for triage in pleural tuberculosis

Triage in a contagious disease as pleural tuberculosis is essential to send the patients to a correct treatment to cure the ailment. These tools are still a challenge because some invasive standard tests are necessary to detect it. The present work shows two models to do a clustering and classify patients of pleural tuberculosis in three risk groups. This clustering uses Fuzzy-Art neural networks to find these groups in the data. First approach employs just anamneses variables and second methodology uses additional information about some classical result tests. The last approach exhibits best results compared to the approach using anamneses variables. Results for sensitivity are similar in two approaches, presenting 93.75% in the anamneses case and 96.87% using more variables.