Abstract:
PredicTB is an intelligent system for the classification of tuberculosis that could help
healthcare providers diagnose and classify patient data, especially for rural areas where doctors
are not always available. The system uses a multilayer feed-forward artificial neural network to
learn and predict patterns in the patient data. PredicTB uses algorithms like backpropagation and
incremental pruning in order to reach a sufficiently accurate diagnosis. A sample project was
created to test how the system performs. The data set was obtained from 174 patients from the
UP Prime TB DOTS. Data was classified into two types: pulmonary and extrapulmonary TB. The
system was able to reach 96.49% accuracy and from these results, we can conclude that an
artificial neural network is an accurate and reliable method for classifying tuberculosis patients.
