Author_Institution :
Siemens Corp. Res. Inc., Princeton, NJ, USA
Abstract :
We present in this paper a thoracic cage coordinate system for recording pathology locations in lung CT volume data. The current recording schemes in a computer aided diagnosis (CAD) system are either only valid for a particular data set or very coarse. We need a coordinate system that is independent of pose and translation, and is relatively stable over deformation and inhalation. The thoracic cage is composed of several pairs of ribs, the spine and the sternum. These features are very stable and prominent in CT data and cover the complete lung area. The rib structures are highly ordered and symmetrical, and each pair of ribs roughly forms a plane. The centerlines of each individual rib are extracted and labeled from top to bottom. For each pair of ribs, a 3D orthogonal basis is computed by eigen-analysis of the rib centerline points, which are taken as the x, y, and z axes. The rib pairs form a set of reference planes and therefore there are a set of coordinate systems (x, y, z), each of them is locally valid between two adjacent planes. To define a location globally, a fourth parameter n is added to identify the serial number of the reference plane. The complete coordinate is recorded as (n, x, y, z). This system is robust against deformations due to bending and twist, and is relatively stable over inhalation. We see this system readily adapted in other 3D modalities, such as MRI volume data.
Keywords :
bone; computerised tomography; lung; medical image processing; 3D orthogonal basis; bending; computer aided diagnosis system; deformations; eigen-analysis; lung CT volume data; pathology locations; ribs; spine; sternum; thoracic cage coordinate system; twisting; Computed tomography; Data mining; Lungs; Magnetic resonance imaging; Pathology; Ribs; Robustness; Software algorithms; Sternum; Telephony;