This paper is devoted to the development and testing of a software package designed for intelligent prediagnosis of digitized computed tomography images of the human lung. A hybrid approach for combining algorithms of digital image processing and analysis is proposed for the development of the software application. In the first step, an algorithm for pre-processing digital images based on filtering and segmentation techniques has been presented. The second step involves the implementation of algorithms for the assessment of scaling characteristics using fractal and multifractal image analysis methods. The classification of computed tomography images was carried out using a convolutional neural network construction and training method. The characteristics of the Renyi spectra for images corresponding to the identified pathologies were analyzed. The classification of the images into the categories corresponding to ‘norm’ and ‘pathology’ on the training set was also performed using the developed techniques.