Semiautomatic registration of the muscles from CT DATA

Computed tomography (CT) scans of muscles appear as poorly contrasted structures. So it has always been a challenging issue for registration. Research focuses on a semiautomatic registration algorithm that supports interactive deformation of a generic atlas model into a patient-specific model. First generate synthetic scans along pre-defined slicing axes in patient space. Muscle outlines obtained by an initial mapping of generic atlas 3D models are then superimposed on these 2D slices. These outlines are deformed by simple geometric operations to match the underlying muscle shapes using a novel software tool. In addition, resolved the intersection problem of multi-interpenetration muscles generated by the above steps. Experimental results demonstrate the high efficiency of the new algorithm.