I'll come back to the issue of 'validating procedures and conducting proficiencies' later.
The main concern I have is that there are X-ray beam geometry effects that aren't being taken into account before the various measurements are done, to arrive at a calibre determination by means of radiography. I have set up a scene in a raytracing program to produce images that may make it easier to understand this phenomenon. The divergent nature of the X-ray beam can be modelled by light rays and I am confident that the images below can be duplicated on an X-ray imaging system. These images demonstrate magnification and beam projection effects when the variables being altered are:
1) Subject to film (or digital recording medium) distance
2) X-ray source to film (or digital recording medium) source
3) Imaging via central ray vs rays not at 90 degrees to the subject and film (or digital recording medium)
I have made all the elements as accurately as I can in the short space of time I have allocated myself to get this out there. Depending on where this leads, I can spend more time on it. Feel free to invite any radiologist or medical physicist to participate, I welcome any criticism of my work here.
I made a 9 x 9 subdivided cube and set it up on a floor, which would be the position of the recording medium (or film). I set up a spotlight, aimed at the center of the cube, 90 degrees to the cube and film. Then I made nine identical spheres, each one fits exactly within any constituent cube of the model. Here is a the model with the spheres lined up in a row, occupying the lowest central row of the cube:
Here is a front view of the setup, zoomed out to see the X-ray source:
This represents the best possible circumstances under which these spheres could be X-rayed, because the central ray of the beam is aimed at the central sphere, the row of spheres is placed directly on the film and the X-ray source is relatively far away from the film, which means that magnification effects are reduced. Note also that this model being presented here assumes a point X-ray source. In reality there is no such thing as a point X-ray source. The source of X-rays is a small area, usually in the order of 1mm squared or less.
I use that cubed matrix to position the spheres, and then I make it invisible for the render, otherwise the image is too busy. I have duplicated the bottom 9 x 9 matrix and left that visible so that you can see where the resultant shadows of the spheres sit on the film, relative to the position of the spheres in the cube.
Here is the render of the best case scenario, from above (viewed from X-ray source towards the film):
The spheres are rendered with a candy texture to make shadows more apparent. Here you can see small shadows projected away from the spheres at the beginning and end of the row.
Next, the spheres were moved away from the film, to occupy the uppermost central row in the cube:
The X-ray source hasn't moved, only the spheres have:
Here is the resultant render:
Note how the shadow is now visible in association with all the spheres. Note also that on the final radiograph, you would not see the candy textured spheres, all you would see would be the shadows (usually in white, not black). See my avatar for an example of a bullet shadow near a person's hip.
There are two critical observations:
1) The X-ray shadows are larger than the spheres. This happens with all X-ray imaging that has not been digitally manipulated.
2) The shadows are not central to each sphere (except for the one that was in the path of the central ray). The shadows are projected away from the spheres in the direction of (and subject to) the oblique rays of the X-ray beam.
The next variable we can change is the X-ray source to film distance (in the US it is commonly called SID, and in the UK and SA it is commonly called FFD). If we reduce the SID as follows:
We get this render:
The peripheral spheres have now been X-rayed by more oblique rays than they were previously. Their shadows are subject to increased beam projection effects. They are more magnified, the shadows are more offset with respect to the actual position of the sphere relative to the grid and even more worrying: the profile of the shadow has changed!
To make it more apparent we can bring the X-ray source in even closer as follows:
And then the render is this:
The beam projection and magnification effects are so gross now, that only 4 of the shadows can be seen in the render view and depending on the size of the film or recording medium, shadows could be lost off the film. In trauma radiography this can result in failure to visualise projectiles that are in situ, in obese patients when the projectiles are anterior and the patient is X-rayed supine (with limited operation SID). Further views would have to be done to ensure all skin margins were included to make sure projectiles weren't being missed.
The effect I want you to take note of is the elongation effect of the sphere shadows, where the spheres were not in the path of the central ray.
Remember this is still a very easy setup. The row of spheres is central to the cube and to the X-ray beam. What about a single sphere, placed somewhere in the cube?
If we go back to maximum SID, here is the render:
This equates to real life circumstances where the position of the sphere is not known. Remember you would not see the sphere, you would only see the shadow.
The next two images show the position of the shadow if the SID is reduced by half, as was done with the row of spheres above.
Do you see what has happened? The shadow of the sphere is projected left AND down, not just in a horizontal direction as before. This is because the sphere is not located centrally in the cube (neither in the X nor z plane). Note also the distortion of the shadow because of elongation effects of the X-ray beam.