I did two WL tests with small BB shift beetwen test, to check if results depend or not of the BB position.
I get this result.
Number of images: 16
Maximum 2D CAX->BB distance: 1.23mm
Median 2D CAX->BB distance: 1.04mm
Gantry 3D isocenter size: 0.69mm
Gantry iso->BB vector: Vector(x=0.07, y=0.00, z=-0.87)
Gantry sag in the z-direction: 0.66mm
Collimator 2D isocenter size: 0.23mm
Collimator 2D iso->BB vector: Vector(x=0.15, y=-1.12, z=0.00)
Couch 2D isocenter size: 1.18mm
Couch 2D iso->BB vector: Vector(x=-0.17, y=-0.64, z=0.00)
Number of images: 16
Maximum 2D CAX->BB distance: 0.83mm
Median 2D CAX->BB distance: 0.67mm
Gantry 3D isocenter size: 0.96mm
Gantry iso->BB vector: Vector(x=0.05, y=0.19, z=-0.35)
Gantry sag in the z-direction: 0.66mm
Collimator 2D isocenter size: 0.31mm
Collimator 2D iso->BB vector: Vector(x=0.14, y=-0.65, z=0.00)
Couch 2D isocenter size: 0.77mm
Couch 2D iso->BB vector: Vector(x=0.11, y=-0.37, z=0.00)
It’s a “normal” difference for you ? Or you think is too large ?
About Gantry 3D isocenter size: 0.96mm // 0.69mm
Couch 2D isocenter size: 1.18mm vs 0.77 mm
Rachid,
I can’t answer definitely since I don’t know your exact setup and what your shift was, but the results look fairly reasonable. The couch iso size had a somewhat large difference however; did you take the two tests with the same conditions? If the CAX and BB locations look reasonable on each image then it is reasonable to believe the values acquired. Note that for very small radiation fields, any small MLC error may introduce relatively large shifts in the CAX location.
I’m working on getting a manuscript in either Med Phys or JACMP that showcases pylinac; in that we benchmarked the major modules against hand measurements or known values. For WL, we compared to MPC and results were within 0.2mm of MPC results. What’s important to know however is that if you take different angles than MPC, you might get larger differences (e.g. couch angles at 90/270).
I can put the BB perfect on Gantry or Couch, I have a micro mechanical device to do it.
About my setup, I only shit aournd mm in each axes…to see the impact.
Ok for MLC point of view !!
I will repeat the test without any shift to see the “test precision (MLC)”
Then, I will do the test with “cone” not brainlab cone but only a home cone to have results independant of MLC precisions.
Could you help me guide me to change your code for “circle” search on images and not square…to run W_L analysis.
This can be useful also for linac with cone accessory. Novalis. It’s not my case, it’s only a home cone for mechanic test nothing else.