Artificial neural network models for diagnosis support of drug and multidrug resistant tuberculosis

Tuberculosis (TB) remains as one of the major health problems in the world. The diagnosis of drug-resistant tuberculosis demands even more attention, leading to longer treatments and higher deceased rates. All methods available to do so have deficiencies in its detection rates, response time, or have a higher cost and need a complex infrastructure to setup. In this work, by using responses retrieved from anamnesis, it is proposed the development of a neural network model to support drug-resistant TB diagnosis. Data are originated from reference centers addressing drug-resistant TB in Rio de Janeiro, comprising exclusively patients presumed to have drug-resistant TB. For the classification between drug-resistant and drug-sensitive TB, the sensitivity and specificity reached 95.15 ± 4.13 and 85.47 ± 5.68, respectively.