Due to the large effort associated with the implementation of UHCF experiments, the development of alternative methods is very important. Therefore, in the present master thesis, a statistical evaluation method is used to predict the distribution of large non-metallic inclusions based on SEM/EDX results. The investigated steel grades are tool steels from Böhler Edelstahl GmbH & Co KG. The materials were manufactured by different production routes. Hot work tool steels, conventionally poured off, pressure electroslag and vacuum remelted, as well as a powder metallurgical produced cold work tool steel grades were compared to each other. In the statistical analysis of the SEM results, the 2D-distrubution distribution of the detected inclusions was transformed into a three-dimensional distribution. The failure inducing defects, found on the fracture surface of giga-cycle fatigue testing, were used to estimate the parameters for an extreme value distribution. This distribution is also transformed into 3D in order to determine the correlation between SEM/EDX-and UHCF tests results. Based on SEM results it is possible to determine the size distribution of non-metallic inclusions of high-purity steels. A significant influence of the production route and the consequent size spectrum of the inclusion on the correlation are observed