Automated model driven localization of the heart and lung surfaces in thoracic MR images

Proposes a representation for anatomical shape knowledge, by which the boundaries of the heart and lungs can be localized automatically in thoracic MR sets. The major organs in the thorax are described in a compact mathematical model by combining a set of fuzzy implicit surfaces by means of Constructive Solid Geometry (CSG). The resulting shape description captures both the single organ shape and their spatial context in a boundary potential function. This model potential allows an automated model-image matching step, in which the model is fitted to an image set of a different subject. The matching method was tested on 15 clinical thoracic MR scans front 13 different subjects. In 13 cases, the lung and heart surfaces were localized within a margin of 10 mm. The matching method is applicable to thoracic MR data irrespective of ifs planar orientation, provided that a major parr of the thorax is present in the image volume.