9th International Workshop on Bone Densitometry, Traverse City, Michigan, September 26-30, 1992. Calcified Tissue International, 52:173, 1993.
An automated technique is presented for the extraction and quantification of vertebral body morphology from lateral spine radiographs. Quantitative measurements of vertebral bodies are important for both the diagnosis and evaluation of the progression of osteoporosis. Conventional quantitative methods involve the identification of landmarks to define the vertebral body perimeter, which can producea significant observer error due to the subjective placement of these landmarks. High quality radiographic films, produced with an optimum technique and film-screen combination, were digitized on a high-performance and high fidelity radiographic film scanner. The vertebral bodies were isolated using a combination of several common image processing techniques: digital image filtering, edge detection, and edge enhancement. The separate vertebrae were characterized by determining specific heights and widths at key positions, as well as the total area of each vertebral body projection. Pattern recognition techniques commonly used in computer vision applications contribute to determining these parameters. We expect improvements in the precision and accuracy of the results due to increasing the number of landmarks applied. Results from a preliminary study will be reported. The most significant impact from utilization of this technique may occur in serial studies, where subtle changes in vertebral morphology would be expected to reflect osteoporotic progression.