By CT I guess you mean a clinical CT rather than a pqCT/micro-CT etc. There is a quite good review on this topic here: 

http://www.intechopen.com/download/pdf/25703

Do you mean density (weight of calcium/volume)? There is a very high degree of association between CT values (previous Hounsfield units) (also called attenuation) and weight of calcium/volume. So CT is ideal for assessing bone density, also called volumetric bone density (vBMD) (g(cm3).

If by BMD you mean areal bone mineral density (aBMD), it is more difficult. aBMD is the (total amount of calcium within a bone volume)/(projected area of the same bone volume), the latter often called region of interest (ROI), you really get 'mass per projected area'. By applying mathematic to this we will see that this is the mathematical product of bone density (weight/volume) and bone thickness. So the aBMD is really a combination of bone density and thickness of the bone. Note that the thickness is the dimension not "seen" by 2-D DXA.

So, by CT, aBMD is (mean CT-values within a volume) x (volume) / (area of the volume as seen on a 'scout view').

As to your question of a protocol: Most CTs have a simple program where you can specify the following: Select all voxels (pixels) with CT density 100-200, then you basically get the volume and mean density of the "internal volume" of cancellous bone and the mathematical product of the two is an estimation of the amount of cancellous bone calcium. If you ask for all voxels with a density > 200 you end up with the cortical bone. Usually you must draw a line around the bone of interest to avoid irrelevant calcification. Please, be aware of the difference between the concepts "internal" and "external" bone volumes, as defined in a paper published in the far ancient days.

If you want to measure the "external" bone density, volume and mass, you need to delineate the bone volume, either by drawing a line or using a circle. The latter is the common method for measuring density of cancellous bone in vertebral bodies. 

You may also require the inclusion of a density phantom (of a gradient of known quantities of hydroxyapatite) in the scanner's field of view to establish a grayscale for the determination of hounsfield units.

If the CT has dual energy capability, the bone density programs are very accurate. As accurate as a specialised BMD unit.

You also can use Rad Ant for microsofy computers. It's free (www.radiantviewer.com/).