Model-based analysis of computed radiographs

Algorithms for computed radiography that allow a computer to recognize anatomy in medical imagery and to identify variations from normal in size, shape and density are discussed. The imagery is obtained from a Philips/Fuji computed radiography system that uses reusable photoluminescent image plates. The approach is general, allowing the use of the same system software for any modelled anatomy. Image-processing techniques are used to extract edges, regions, vertices and other relevant image features. These features trigger formation of hypotheses of the imaged bones and soft tissues. Multiple hypotheses of the size and orientation of the imaged anatomy are matched against stored 3-D models of the relevant anatomy, obtained from statistically valid population studies. A typical processing example from a normal hand is shown