The stent of this bronchial model can be seen in the 3D image.
PARAMETERS:
SDD=842mm, SID=455mm. image size=768×768, pixelsize=0.56×0.56. numbers of image=889.
The code I use is :
.\rtksimulatedgeometry.exe -n 889 --sdd 842 --sid 455 -o “C:--------\0100.xml”
.\rtkfdk.exe -g “C:---------------\0100.xml” -o “C:---------\fdk0004.mha” -p "C:------------" -r “.*.dcm” -v “on” --neworigin “-215.32,215.32,0” --newdirection ‘‘1,0,0,0,-1,0,0,0,1’’ --newspacing “0.56,0.56,1” --hardware “cuda” --spacing 0.2 --dimension 1024
Any problem with my code?
This CBCT system is precisely configured, with three Euler angles set to 0 by default. QUALITY of 3D reconstruction is high when using the reconstruction software provided by the manufacturer.
I also use calibration parameters for 3D reconstruction in RTK, but the quality did not become higher. In calibration parameters, --out_angle 2.026459e-01 --in_angle -1.082085e-01,… , reprojection errors is 0.0908.
the new code: .\rtkfdk.exe -g “C:------------.xml” -o “C:---------------.mha” -p "C:--------------" -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 0.303 --dimension 512 --radius=2 --multiplier=1.5 --hann=0.7 --lowmem “on”
the new result is better, but there are still artifacts:
The geometry is generally calibrated by the manufacturer. If they have a high quality reconstruction, you probably need to find the geometric parameters in their software. If you can’t find it, you’ll probably have to calibrate it yourself using e.g. ball bearings in the field of view.
In fact, their reconstruction software only requires the input of two calibration parameters, SDD SID, and they gave these two values. I also calibrated myself using a calibration phantom with metal balls and obtained very close results. But their steps and parameter selection are confidential, such as image preprocessing and processing of reconstructed images.
