I’m working on 3D reconstruction using cone-beam X-ray projection images with the FDK algorithm in RTK (Reconstruction Toolkit). After reconstruction, I noticed that some voxel values in the output volume are negative.
What I reconstruct is a head phantom.
I didn’t use -Hann -radius -multiplier .
my code is : .\rtkfdk.exe -g -o -p -r “.*.dcm” -v “on” --neworigin “-214.76,214.76,0” --newdirection “1,0,0,0,-1,0,0,0,1” --hardware “cuda” --newspacing “0.56,0.56,1” --spacing 0.28 --dimension 240 --origin “-40,-80,-30”
and
.\rtkfdk.exe -g -o -p -r “.*.dcm” -v “on” --neworigin “-214.76,214.76,0” --newdirection ‘1,0,0,0,-1,0,0,0,1’ --newspacing “0.56,0.56,1” --hardware “cuda” --spacing 0.28 --dimension 560
It’s not unusual to have negative values as FDK does not enforce positivity. For example, noisy measurements in air oscillate around zero.
I don’t know why the head and vertebra phantoms seem have negative values without looking at the projection images, you should probably check for negative values in them or truncation artefacts.
1 Like
thank you for your help.
The voxel values of my input 2D images are 0 for black (phantom) and 255 for white(nothing or air). Do I need to inverse it?
Yes, it would make sense as the linear attenuation coefficient of air is ~0 and positive for tissues. Are you using RTK to forward project your phantom?