Abstract: A controversial part of the state-of-the-art in deep learning applied to medical imaging concerns the selection of 2D, 2.5D or 3D representations. This study contributes by contrasting the ...