DISCRIMINABILITY OF FRACTURE AND NONFRACTURE CASES BASED ON THE SPATIAL-DISTRIBUTION OF SPINAL BONE-MINERAL

Purpose: The purpose of our study is to demonstrate that spinal minera l distribution measured with CT can distinguish normal from osteoporot ic individuals. Method: CT studies of lumbar vertebrae (L1-L3) from 12 1 clinically normal women without fractures and 57 women with one or m ore atraumatic fractures somewhere in the skeleton were evaluated with discriminant analysis based on indices of the spacial distribution an d noise properties of spinal bone mineral density (BMD). Results: The use of discriminant analysis for all of the normal and osteoporotic wo men (L1-L3) resulted in a classification accuracy of 87.1% for fractur e cases and 83.2% for nonfracture cases. In contrast, using the conven tional method in the same patient population, 62.5% of BMD values of o steoporotics overlapped with those of normals whose BMD was below the 90th centile of osteoporotics. Conclusions: CT-based measures of the s pinal mineral distribution can increase the accuracy of discriminating fracture and nonfracture cases almost to 90% accuracy, even in a regi on below the fracture threshold. This shows that in this region the ri sk of fracture is not completely random but has a stochastic component as well.