We have investigated the importance of various lung characteristics extracted from computer tomography (CT) in combination with other factors for the outcome of the endobronchial valve (BV) therapy in patients with destructive pulmonary pathology due to tuberculosis. We have devised prognostic models on the basis of the decision tree method implemented in R programming language. This allowed us to reveal key characteristics and interactions between them as well as to estimate the success of the endobronchial valve treatment. These models make predictions for changes in the lobe volume and cavity closure at BV therapy. We have found that pleural integrity is a good predictor of lobe volume reduction after BV treatment as well as lobe volume reduction during BV therapy is a prognostic marker for cavity's closure.