User question

[satzhan]

Hello there,

I'm trying to see if I could use STIR to reconstruct image from the simulated data for a bit of custom design PET. It has two layers but they are not nicely stacked and not of the same shape (number of crystals). I was going through the documentation, but I am not sure if we could define detectors for each layer? I saw 'Generic' example with a list of detector coordinates, so I'm guessing if I could use 1 crystal per detector, that is, just listing crystals, but I wanted to double check this here.

Best,
satzhan

[KrisThielemans]

The generic code was put in by people from ETH for their preclinical scanner. It was single-layer though. I'm fairly sure that providing a list of crystals would work as far as projections goes (the ray-tracing matrix might need switching symmetries off, but it likely does that already). The code needs a 2D array I believe, which implies some restrictions. It could possibly be "faked" but having a 1D array.

However, normalisation estimation is not available for "generic" scanners (possibly the "crystal efficiencies" code could be made to work, but not the geometric one as it needs symmetry).

(Object) scatter estimation for the generic case would only work if not down-sampling the scanner at all, as we don't know how to upsample it, which will make it very slow.