Neural network technologies in prediction the effectiveness of treatment of patients with pulmonary tuberculosis

Abstract

The study used predicting the effectiveness of treatment of patients with pulmonary tuberculosis using neural network technologies. The most optimal neural network model was obtained, which allows predicting the effectiveness of treatment with a forecast accuracy of at least 78.4%. As a result of constructing a neural network model, the most significant «input» parameters of the neural network were identified: the presence of hepatotoxic reactions, the level of IL-1ß, IL-6, IL-4, IL-10, IFN-γ, C-reactive protein before the start of the intensive phase of chemotherapy, the presence of antibiotic resistance, the presence of mycobacterium tuberculosis before the appointment of a specific chemotherapy by seeding, the volume of lung tissue damage, the chemotherapy regimen, the clinical form of pulmonary tuberculosis, as well as the genotype of ЕЕ gene GSTT1.