Purpose: Dual-energy x-ray absorptiometry (DXA) systems present semi-automated analyses of lumbar spine and femur bone mineral density (BMD). Complete analysis, however, should include a visual assessment by trained personnel to a) guarantee proper acquisition mode and patient positioning, b) identify regions of interest (ROI) accurately, and c) identify abnormal conditions that could artificially elevate BMD. New software known as Computer Assisted Densitometry (CAD) (GE Medical Systems Lunar) was recently introduced to identify scans with specific, common irregularities introduced during acquisition or analysis. Abnormalities include problems related to positioning, accurate ROI placement, unusual high-density areas, unusual T-scores, and spinal curvature. We compared the CAD assessment with that of two experienced DXA users.
Methods and Materials: Seventy-one spine and 70 femur scans were evaluated for abnormalities in six areas: spine centering, spine alignment, ROI accuracy, unusual high-density bone, unusual T-scores, and spine curvature. Two-thirds of the scans exhibited at least one abnormal condition that could be identified by CAD. Two experienced users with ISCD certification examined the scans and agreed on their findings prior to knowledge of CAD results. CAD results were then compared with their visual assessment. Differences between CAD and visual assessment prompted a secondary review to determine whether a scan abnormality might have been missed by the densitometry team.
Results: Results showed strong agreement between CAD classification and visual assessment in all six areas. Experienced users agreed with CAD assessment in 76% to 86% of cases prior to knowing the CAD results. The densitometry team changed their assessment in 20% to 40% of the cases where there was a disagreement between CAD and visual assessment. Agreement increased to 83% to 92% after CAD assessment was known. Femur results were similar.
Conclusion: We conclude that CAD provides valuable information for inexperienced users regarding DXA scan quality and assists experienced densitometrists in identifying abnormal scans which might have been missed by visual assessment alone. (E.N.S., D.M.S. received grants from GE Medical Systems Lunar.)