||Deformable image registration is an important part of medical image analysis. Applications include longitudinal studies, population modeling, and atlas based image segmentation. However, the registration problem is a hard one and there is no one-size-fits-all solution. Imiomics is a whole-body image analysis concept combining imaging data with non-imaging data to perform various analyses. Imiomics requires a point-to-point correspondence between the images, therefore image registration is essential. A complete registration pipeline for whole-body fat-water MRI images, where bone, water, and fat are registered sequentially is presented. The core part of the pipeline being the non-parametric deformable registration algorithm based on graph-cuts. Together with a number of optimization strategies, including a block-based approach, the method is shown to produce a good result in a reasonable amount of time.