Text this: Multi-Scale Contextual Coding for Human-Machine Vision of Volumetric Medical Images