COVID-19 is an infectious disease caused by the novel coronavirus (SARS-COV-2). The global number of cases and deaths totaled 587 million and 6.4 million in August 2022. As this disease is highly contagious, the diagnosis must be made at its early stage to take necessary measures, including isolating the patient. In addition to the diagnosis, it is also essential to identify how this disease presents itself in patients, observing the involvement of the lungs. With this, it is possible to follow the evolution of this disease in the patient. Thus, we present a segmentation approach using U-Net, combined with a pre-processing and data augmentation methodology. In this work, we used several experiments where the K-Fold represents the best evaluation method. The methodology proposed in this experiment obtained the following metrics: 78.40% ± 0.05 from Dice, 64.80% ± 0.07 IoU, 78% ± 0.07 Sensitivity, and 100% ± 0.00 Specificity.