hide
Free keywords:
-
Abstract:
A system and method (12) for determining a property map (82) for an object, in particular for a living being based on at least one first image (84) in particular a nuclear magnetic resonance image, an MR image, of the object. According to the method (12), a structure for reference pairs is defined in a first step (96), each reference pair (16 - 26), having at least two entries (62). The first entry represents a property value, in particular, an attenuation value. The second entry (62) preferably represents a group of related image points (67), in particular, extracted from MR images (28) comprising an image point of interest corresponding to the property value. A number of training pairs (16 - 26) are prepared in a further step (98) of the method (12). A structure for the training pairs (16 - 26) corresponds to the structure of the reference pairs and the entries for the training pairs (16 - 26) are known. A relationship between the first entries and the further entries (62 - 66) of the training pairs (16 - 26) is established by machine learning in a further step (100) of the method (12) in order to be able to predict a corresponding property value (88) for any point (90) of the first image (84).