<p>Improving liquid injection by enhancing breakup capabilities is vital to face climate change for industries that still require combustion processes. In this work, a numerical methodology is developed to filter droplet population results in DNS primary atomization simulations focused on aeronautical pressure-swirl atomizers when using mass-conservative formulations. The main goal is to improve the reliability of the outcomes so that prediction models can be elaborated. A filter based on the volume of the liquid structures detected is developed and implemented in the DNS code’s droplet detection algorithm. Experimental results are used to test its effectiveness. After filtering, DNS droplet size Probability Density Functions are considerably improved, getting distributions and average diameters closer to the experimental ones. Axial velocity distributions show the same trend as experiments but, generally, the DNS model underpredicts negative velocities, probably due to the limitations of the one-way coupling approach when feeding inflow boundary conditions. This study means a step forward in primary atomization prediction through high-resolution methods. Its results can help to get more accurate statistics in terms of characteristic sizes and droplet size and velocity distributions, which are widely used in LES and RANS calculations to model primary atomization, leading to an increase in the temporal efficiency of the research process.</p>