How are Dual-energy X-ray absorptiometry (DXA/DEXA) scans affected by surface stability? Today I had my body scanned by a GE Lunar Prodigy densitometer installed in a small RV of a company that tours California offering body composition scans (image from Wikipedia article on DXA).

During the scan, I felt the surface wiggling regularly about 1-2 centimeters left and right, as the scanning arm was moving. When I descended form scanning bed, I could feel the floor of the RV move easily, due to the truck's suspension system. I tried to move the scanning bed by hand but it wouldn't budge.
How would this sort of placement of a densitometer on struts affect scanning accuracy?
(If this is a better fit for PF or Health.SE, please migrate.)
 A: When talking about accuracy we have to keep in mind, that we have to distinguish between the accuracy of the material decomposition and the accuracy of stitching the single images of the scan to a big one.
Probably the most important one is the quality of the material decomposition. It is done by taking alternating images with different X-ray spectra. This can be done by changing the kV of the tube and/or the used filtration. One spectrum will have more lower energetic X-rays and the other one more higher energetic ones (I can add spectra if someone is interested in that). The material decomposition can now be calculated by using the different energy dependencies of the X-ray absorption of high Z and low Z materials. The dependencies are mainly driven by the photo- and Compton effect.
As the switching between high and low energy happens on a frame to frame basis slow movements are not critical for the material decomposition. However the borders of regions with strong contrasts (bone to soft tissue) will get blurred out.
The stitching of the individual images however might look awkward, when you moved during the procedure. But as the image integrity of the 2D image is probably not of clinical relevance this should not be an issue.
A point I did not take into account is the impact of the relative movement of source and detector on the calibration of the detector (gain correction).
The separation Ca and the rest is the easy part, as Ca has Z=20 and the rest are low Z materials O(Z=8), C(Z=6) and H (Z=1). As there are only 2 independent parameter you will effectively do a material decomposition in a basis of Ca and soft tissue. Without further information it is not possible to distinguish between water and fat. This is easy to see when you look at the absorption coefficients of the different material. The low energy image might measure the average absorption in the range of ~30-70keV and the high energy image the part from ~70-130keV. These are rough estimations, the real values will be dependent on the kV setting of the tube and the used filters.

A: Half a year later, I shelled out $90 and scanned myself two days in a row, at the same time of the day, fasting overnight each time, using the same GE Prodigy machine in the Body-Spec truck. Turns out I had lost 1.8lbs of lean tissue overnight!

I blogged about how (in)accurate DXA was and would love to find out what I can do to minimize error, or if others can replicate my results.
