The present disclosure relates to a method and a system for modelling a human heart and/or chambers and cavities therein, such as the left and right atrium, based on a plurality of emission tomography images, such as positron emission images, of said human heart, wherein each image represents concentrations of a tracer that has been injected at a specific time, the overall method comprising the steps of: A: extracting time activity curves of a tracer that has been injected for a plurality of pixels and/or voxels； B: identifying first-pass peaks of the time activity curves, each peak corresponding to an arrival time of the injected tracer at the corresponding pixel/voxel； C: defining a model comprising at least two portions of the heart, wherein the at least two portions are isolated by selecting pixels and/or voxels based on comparisons of the first-pass peaks against a threshold； and D: arranging the at least two portions in relation to each other by comparing the arrival times of the first-pass peaks of the pixels and/or voxels in the portions, thereby obtaining a segmented model of the human heart, which method and model may specifically be used to obtain an estimate of the volume of the left or right atrium of said human heart.
