Learning Disentangled Representations in the Imaging Domain

•Tutorial about disentangled representation learning, i.e. learning how to separate out the underlying factors of variation.•Key concepts of machine learning and learning representations, causality, and domain shifts.•Detailed discussion of models that enforce disentanglement, their building blocks, and the existing metrics.•Comprehensive survey of disentanglement applications in medical domain•Analysis of disentanglement applications in medical imaging, discussing biases, models, and training setups.•A comprehensive listing of limitations, open challenges and existing opportunities in the field of disentanglement. [Display omitted] Disentangled representation learning has been proposed as an approach to learning general representations even in the absence of, or with limited, supervision. A good general representation can be fine-tuned for new target tasks using modest amounts of data, or used directly in unseen domains achieving remarkable performance in the corresponding task. This alleviation of the data and annotation requirements offers tantalising prospects for applications in computer vision and healthcare. In this tutorial paper, we motivate the need for disentangled representations, revisit key concepts, and describe practical building blocks and criteria for learning such representations. We survey applications in medical imaging emphasising choices made in exemplar key works, and then discuss links to computer vision applications. We conclude by presenting limitations, challenges, and opportunities.