Tuberculosis (TB) is a global public health problem and thousands of people still fall ill and die from this disease and its complications. Taking this concern into account, this work aims at using artificial neural networks based on the Perceptron model to identify the presence of tuberculosis in pulmonary radiographs. For that, a database with 892 images was consolidated, containing several manifestations of TB as well as healthy examples. The results obtained by artificial neural networks produced good predictors for this task, with few neurons in hidden layers and F -Score above 0.72, consolidating an strategy to support TB diagnosis.