Multifractal analysis provides a different way to process images. Instead of assuming smoothness properties, one characterizes the different elements of an image (e.g. edges, textures) by their regularity. This regularity is measured both in a local way, through the Hölder exponent, and in a global one, through the multifractal spectrum. Such an approach is particularly fitted for medical images, which are by their very nature mostly irregular. We provide examples of applications of this new paradigm to various cases.