A Novel Automatic Framework for Scoliosis X-Ray Image Retrieval

Abstract

The paper proposed a novel automatic scoliosis X-ray image retrieval framework based on the global statistical feature of edge, edge co-occurrence matrix (ECM) and the local geometrical feature set of the whole spine, angle of each spine curve. The ECM is based on the statistical feature attained from the edge detection operators which applied on the image. The eigenvectors obtained from principle component analysis (PCA) of the ECM can preserve the high spatial frequencies components, so they are well suited for shape as well as texture representation. The geometrical feature like the Cobb's angle of each spine curve could be derived from the image segmentation based on the Intersecting Cortical Model, which is elicitation of the Eckhorn's model. The experiment shows that the framework shows good accuracy for the input query X-ray image in our work.