Semi-automatic detection of fracture in wrist bones using graph-based grammar approach

Medical image analysis is systematic evaluation which is similar to expert knowledge of radiologists and other clinical specialists. To perform this, cognitive recognition facilitates the semantic analysis and understanding of the images. In this paper, the X-ray images of wrist bones are analyzed for fracture identification in metacarpal region, using graph based grammar approach. For a normal wrist X-ray image, graph is constructed by considering the ulna of the wrist bone as the starting node and the graph is traversed through each bone which are assumed as node and the path is ended when it reaches the phalanges. From the constructed graph, the production rules are generated from the ulna(start symbol). For the test image using the grammar rules, input strings are generated and validated using the parser. If a valid strings(normal X-ray) are given, it reaches the final state, else it reaches an invalid state(fractured image). Based on the state it reaches, the type of the fracture that exist in the wrist X-ray image is identified. Any kind of fracture in the wrist bone can be easily detected by extending or adding the production rules without modifying the system is an important aspect of image understanding.