Semi-Auto Lab Experament

This week in our imaging lab, we decided to look into the EDR functions even further by performing some experaments on our phantom using the Semi-Auto option, which selects region 5 on the FUJI CR image plate. Our control image, shown first, came out exactly like you would expect it to having the tube, part and bucky aligned, and it was processed on the default "auto" setting (all images in this experament utilized a technique of 70 kVp 3 mAs and 40" SID in the table bucky):



Image #2 kept all of the same alignment, technical factors, and positioning, but we processed it on "semi-auto" to see if the image would come out better... in my opinion on this phantom, I didn't necessarily like it as much. The bony detail through the petrous ridges is lacking and it has somewhat of a longer scale of contrast - undesired if performing this view for sinuses. Note the slight change, yet acceptable S# differences between these first two exposures.



For exposure 3, we decided to keep the semi-auto setting, but we took the x-ray tube off of transverse detent so that everything was aligned except the tube. This is how the field light appeared over the phantom with the cassette alignment underneath:



As you can see with the resulting image, the S# is 3820, which is severely underexposed. The computer is attempting to adjust the density over region 5 on the image plate. Over half of region 5 lies outside the exposure field, which makes the computer think "there aren't many photons here, so it must be underexposed... I'll try to darken the image to compensate because I'm a super-smart computer." We know coputers aren't that smart, and the result is a very dark image over the small bit of anatomy we can see due to automatic rescaling, and it retains its very high S#.



For our 4th image, we kept it on semi-auto, but we started fresh with everything aligned... we then moved the phantom out of alignment. The tube and bucky were aligned but the skull was off-centered like so:



In this image, the S# is 50, which overexposed. The computer is attempting to adjust the density over region 5 on the image plate. Most of region 5 is exposed with the full technique appled without any absorption by the tissue of the phantom, which makes the computer think "there are too many photons here, so it must be overexposed... I'll try to lighten up the image to compensate because I'm a super-smart computer." Again, not smart, just playing by a set of rules and the result is a very light image over the small bit of anatomy we can see due to automatic rescaling, and it retains its very low S#.




We thought these errors were so interesting that we decided to try the alignment mishaps on the "auto" setting to see if it would be more forgiving, or if we would yield similar results. Image 5 was taken with the same misalignment of the x-ray tube that occurred in image 3, but on "auto." Below we see the S# slightly higher than the optimal range at 308, but certainly looking better than exposure 3.



And finally, we reproduced the off-centered patient as in exposure 4 on "auto" and produced a low S# of 50, but also looking better on the auto setting than on the semi-auto where field 5 was the main point of interest.



The lesson learned is that we must out-smart our computer software... be extremely careful with positioning and alignment when selecting any EDR option that will only form a histogram based on a selected region.