OpenKBP - 2020 AAPM Grand Challenge Forum

Go back to competition Back to thread list Post in this thread

> where are the beamlets ?

can we know the position of the beam and compute a coarse dose matrix ? how far is the beam from ptv ? and where is the isocenter

Posted by: argman @ April 9, 2020, 8:19 a.m.

if we compute a coarse dose use CERR, without further optimization, does it violate the rules?

Posted by: argman @ April 9, 2020, 10:34 a.m.

I answered all your questions below, but I'm unfamiliar with the term "coarse dose matrix". Is it equivalent to what we called a dose deposition matrix in the challenge documentation? That is, a matrix whose elements represent the dose delivered to a voxel per unit of beamlet (i.e., small portion of the beam) intensity.

All plans were delivered via nine equidistant beams positioned at 0˚, 40˚,... 320˚. The parameters of those beams were calculated using mostly default settings in CERR, but I can't provide more details because I have no access to the code right now, however, I believe the beams are all 1m from the patient isocentre.

You may use CERR (or any other software/code base) to help engineer new features, and you are allowed to do "further optimization" if you can generate the necessary data. However, you will be scored on the same metrics as everyone else, which means you will be penalized for doing "better" than our ground truth plans.

Posted by: OpenKBP @ April 9, 2020, 3:38 p.m.

And about target area, in the dataset , if there exists ptv56, ptv63 and ptv70, do you choose all these structures as targets or only PTV70?

Posted by: argman @ April 10, 2020, 2:56 a.m.

All PTVs that exist in each patient are classified as targets. For example, the target area for some patients might be PTV56 + PTV70 and for other patients it might be PTV56+PTV63+PTV70

Posted by: OpenKBP @ April 10, 2020, 3:01 a.m.

If some OARs are missing, how do you set the optimization rules(like Dmean, D95) ?

Posted by: argman @ April 10, 2020, 4:04 a.m.

That's a modelling choice that you need to make. Personally, I would just make a separate optimization model for each patient, so if that patient is missing an OAR or PTV you won't need to make an objective for it.

Posted by: OpenKBP @ April 10, 2020, 4:27 p.m.

how do you compute the isocenter ? is it the center of the image or center of targets(PTV) ?

Posted by: argman @ April 19, 2020, 10:21 a.m.

I can't provide the exact isocenters, but the center of the image should be a good approximation for the isocenter that we used for planning.

Posted by: OpenKBP @ April 20, 2020, 6:40 p.m.
Post in this thread