In the article we analyze computer-aided tomography data of patients with confirmed diagnosis and spherical masses found in lungs. An average density (in Hounsfields) of the area of interest, a density standard deviation, a fractal dimension, and a nonlinearity factor are proposed as prospective parameters for differential diagnostics of diseases. The fractal dimensions of medical images are evaluated by the most appropriate Hurst’s method for estimating fractal dimensions of small size images. We then obtain distribution functions of the proposed parameters for two types of diseases (lung cancer and tuberculosis). Also, the prospectiveness of the proposed parameters for differential diagnostics is investigated. In the conclusion, we demonstrate that due to statistic nature of results obtained it is necessary to increase number of diagnostic parameters.