a bug in sequencing2ApertureInfo and some question

[Luliny]

I am just starting to use MatRad and have got a lot of help from the discussion. I want to report a possible bug if it may benefit anyone:

in the matRad_sequencing2ApertureInfo function: in the line: lim_l(l) = (lim_lInd-1)bixelWidth + minX - 1/2bixelWidth;
Matlab will report error if the Lim_Lind is nan.

Also, I have some questions: 1. when using MatRat photon optimization with VMAT, is there a way to choose a specific collimator angle? 2. Is there any development work going on to add multiple VMAT arcs functions to VMAT plan?

[wahln]

Thanks for reporting this, I have to figure out again why there are NaN's in the first place and how we deal with them. It might come from a setting where matRad creates beamlet ray setups with holes in it, especially on the VMAT branch I am currently not sure if this is consistent.

Regarding this: The VMAT is an ongoing implementation documented in PR #377. This PR has been there for a loooong time, but it heavily interfers with matRad's core, so it has been continuously pushed back. It "works" in normal applications, but there are a few edge cases that we are still figuring out.

Thus we are happy for any help, bug reports, improvement suggestions we can get on this. am currently trying to irregular spend some time on sanitizing the VMAT implementation. While doing this, I am also trying to make it more flexible .

Regarding the collimator angle:
There was some work going into collimator angles. For VMAT I don't think it is currently possible, but I can keep it in mind.
It however depends on how we want to set this up. Should the collimator rotation be chosen upfront or during sequencing? If upfront, we can just rotate the beamlet grid and not touch sequencing. If chosen during sequencing, we need figure out a strategy on how to choos rotation and how to properly do some fluence image resampling on a rotated grid first.
What do you think is more appropriate?

Regarding the multiple arcs:
This is not included in the implementation for now, I think. However, as documented in the comment here, #377 (comment), I started pairing angles with arc indices while making the definition more consistent with default matRad style as used for IMRT definition.

Happy to have someone like you trying this out and discuss with us where the stuff fails, this would help us in iteratively updating this PR.

[Luliny]

Thank @wahln for replying! Appreciate your work.

The lim_lInd get nan when the entire row of ray is closed.
Regarding collimator angle: the current implementation of VMAT only allow fixed collimator angle in one arc, so it can be chosen upfront. Unless you want to implement the Varian RAD (rapidarc dynamic), which will allow collimator rotation during on arc.