Precision of dual photon absorptiometry measurements: comparison of three different methods of selection of the region of interest.

To determine the short-term reproducibility of bone mass calculations with dual photon absorptiometry of the lumbar spine (L2-L4), duplicate measurements in healthy subjects were used. Three different methods for selection of region of interest were compared: a rectangular region with variable height and width, an irregular region to be drawn freely by the operator and the standard calculation software supplied with the bone densitometer. Contributions of changes in size and location of the region of interest on calculated bone mineral mass were also investigated. An increase in height of the region by 2 scan lines caused an increase in bone mineral content (BMC) of 8.4% +/- 1.8%. Enlargement in width by 2 pixels in each scan line caused an increase in BMC of 2.5% +/- 1.3%. The difference between these region of interest changes was significant (P less than 0.0005). The use of a rectangular region, optimized to enclose L2-L4 in each person but of the same size in both measurements, resulted in a reproducibility of 1.4% and 0.7% respectively for 2 observers, with an inter observer variation of 1.2%. The reproducibility of the duplicate measurements was worse for the other methods of region of interest selection. A further series of duplicate DPA measurements in normal subjects, but with a standard meal between the measurements, showed a larger variation in the results. Again the reproducibility of the calculations using the rectangular region was better than with the other methods, with much less inter observer variation.(ABSTRACT TRUNCATED AT 250 WORDS)