There is a growing research interest in reliable content-based image retrieval (CBIR) techniques specialized for biomedical image retrieval. Applicable feature representation and similarity algorithms have to balance conflicting goals of efficient and effective retrieval while allowing queries on important and often subtle biomedical features. In collection of digitized X-rays of the spine, such as that from the second National Health and Nutrition Examination Survey (NHANES II) maintained by the National Library of Medicine, a typical user may be interested in only a small region of the vertebral boundary pertinent to the pathology: for this experiment, the Anterior Osteophyte (AO). A previous experiment in pathology-based retrieval using partial shape matching (PSM) on a subset from the above collection; about 89% normal vertebrae were correctly retrieved. In contrast only 45% of moderate and severe cases were correctly retrieved, and on the average only 46% of the pathology classes were correctly determined. Further analysis revealed that mere shape matching is insufficient for semantically correct retrieval of pathological cases. This paper describes an automatic 9 point localization algorithm that incorporates reasoning about boundary semantics equivalent to that applied by the content-expert as a step in our enhancements to PSM, and results from initial experiments.