The present invention relates to a method (200) for iterative material decomposition of multispectral image data. It is described to: a) provide (210) a plurality of spectral images of a region of interest comprising a body part； b) provide (220) a plurality of sets of material coefficients for a plurality of materials, wherein for a set of material coefficients each material coefficient is associated with a corresponding material, and wherein each set of material coefficients is associated with a corresponding spectral image of the plurality of spectral images； c) decompose (230) each spectral image of the plurality of spectral images into a plurality of material images and an offset image, wherein different material images correspond to different materials of the plurality of materials, and wherein a material image is represented by the material coefficient for the corresponding material multiplied by material concentrations at different image locations, and wherein the material coefficient is one of the material coefficients from the set of material coefficient for the corresponding spectral image； e) manipulate (240) at least one of the material images for each spectral image on the basis of at least one topological constraint relating to the body part to determine for each spectral image an updated plurality of material images and an offset image； f) recompose (250) a plurality of spectral images, wherein each recomposed spectral image is recomposed from the corresponding updated plurality of material images and the offset image； g) compare (260) intensities at image locations in a recomposed spectral image to the intensities at the image locations in its corresponding spectral image prior to recomposition to determine a plurality of corrections, wherein a correction is associated with an image location； i) modify (270) the updated plurality of material images for each spectral image comprising utilization of the corresponding plurality of corrections； j) itera
