Since the data space grows exponentially with the number of features, the number of samples required to approximately cover its volume soon exceeds what is feasable in real-world applications. This phenomenon is known as the “Curse of Dimensionality”. Medical data in particular is usually high-dimensional and even multi-modal. Additionally, gathering of medical data is difficult due to the sensitive nature. Autoencoders can learn low-dimensional representations of high-dimensional data. This opens up interesting possibilities to reduce data requirements or even to reconstruct missing modalities.