 This paper presents a novel approach for detecting tuberculosis, TB, from chest x-ray images. It uses a combination of pre-trained deep learning models and handcrafted features to extract relevant information from the image. The extracted features are then optimized using a firefly algorithm and ranked according to their relevance. Finally, the features are combined and used as input to a decision tree classifier which achieves high accuracy, 95.2% and 99%. This article was authored by Venkatesan Rajinikanth, Sifadine Kadri, and Pablo Morenoger.