The main contribution consists of a deep learning model that can identify CT images affected by COVID-19 without the need of a sophisticated task to make it work. For this purpose, a modular architecture is developed to work with 3D data, taking advantage of the correlated information between cuts when deciding.