New contour point model for retrieval of cervical spine images

摘要

Our objective is to develop new geometric eight-point model for cervical spine vertebra that provides sufficient information for diagnosis. A subset of cervical spine images from National Institutes of Health digitised X-rays database is selected for our study. As angles at anterior corners of spine vertebra appear meaningful features for detection of anterior osteophytes, three variants of the geometric eight-point model are proposed to find the best estimate of these angles. Algorithms have been proposed to automatically locate the said eight contour points. The proposed models are uniform for all (normal and abnormal) vertebrae, not biased by human experts and thus free from ambiguity. The angle-based features derived using the proposed models discriminate the normal and abnormal spine vertebrae reasonably well. These models can be used to represent the spine vertebrae shape in lower dimensional feature space and may be useful for indexing and retrieval of cervical spine images.