Multi isocenter winston lutz

Hello everyone,

Does anyone know if pylinac could work for a multi isocenter winston lutz cube? I was thinking maybe you could break up the overall images so that each BB has it’s own set of X images and then run the code once for each BB, but does anyone have better ideas?


Do the images have all BBs in view at once, or are they separate images with the jaws/mlcs collimated around the other BBs?

Thanks for your response James.

We haven’t gotten that far yet. One option, that would be faster to acquire, would be to take one image that has all the BBs in one image with the MLC collimated around each BB separately. I guess another option would be, if you had X BBs, to take X images per gantry/collimator/couch combination where in each image only one BB is visible. I guess the latter case would work properly in pylinac, right? Might be tedious, though to acquire so many images. Any other ideas?

As far as pylinac analysis goes, one image per BB is easiest, but that’s the opposite for efficiency. No doubt the user would want to just take the same number of images they always have and just open up or collimate about the MLCs to see more BBs. The problem here is that it’s nearly impossible to track the BBs emperically when you start moving the axes, especially the couch, unless the arrangement is optimized for this. Note also it’s difficult to not have BB overlaps (or at least adjacency) at some axis combination. The user would very likely have to provide the geometric positions of the BBs a priori so that the expected position could be known. You can see this with a tool like MPC where there are windows of the expected position. I would choose MPC over any other commercial multi-BB solution 24/7/365, but presumably, not everyone has a TrueBeam =). In theory, this could be done by passing a tuple of the x/y/z relative to one of the BBs or CAX. E.g. .analyze(…, secondary_bb_positions=((-5, 15, 10), (-15, 5, 20))). Even so, that’s still a lot of math to keep track of but it should be in the realm of possiblity. Given the new image_generator module though, I should be able to generate some images to test it out. MPC is Varian’s solution, and I know SNC has a tool as well for their own version. Unsure about other vendors and approaches. Open to ideas!