The article depicts an original approach to creation of machine-oriented medical datasets for validation of diagnostic systems based on artificial intelligence. This approach consists of four steps: planning, selection of source data, annotation and verification, and documenting. Examples of medical imaging datasets created in accordance with this approach are presented. The methodology is believed to be universal and scalable, therefore it could be applied to other medical fields other than radiology in which artificial intelligence and big data technologies could potentially benefit medical specialists.

High‐energy accelerators are often used in oncological practice, but the information on the small‐field dosimetry for the photon beams with nominal energy above 10 MV is limited. The goal of the present work was to determine the values of the output correction factor ( ) for solid‐state detectors (Diode E, PTW 60017; microDiamond, PTW 60019), EBT3 film, and ionization chambers (Semiflex, PTW 31010; Semiflex 3D, PTW 31021; PinPoint, PTW 31015; PinPoint 3D, PTW 31016) in the small fields formed by 10, 15, 18, and 20 MV photon beams. The output correction factors were calculated by Monte‐Carlo method using EGSnrc toolkit for six field sizes (from to ) for isocentric and constant source‐to‐surface distance (SSD) techniques. The decrease in the field size led to an increase in for ionization chambers, while for solid‐state detectors and radiochromic film, were less than unity at the smallest field size. A larger sensitive volume of ionization chamber corresponded to a stronger deviation of output correction factor from unity: 1.847 (125 mm 3 PTW 31010) versus up to 1.183 (16 mm 3 PTW 31016) at the smallest field of 10 MV beam. The calculated output correction factors were used to correct the output factors for PTW 60017, PTW 60019, and EBT3. The deviation of the corrected output factor from the results of Monte‐Carlo simulation did not exceed 3% in the fields from to for 10 and 18 MV beams. Thus, Diode E, microDiamond, and EBT3 film can be recommended for small‐field dosimetry of high‐energy photons.