jacg
commented
2 years ago The steel's contribution to the attenuation is proportional to the intersection length between the LOR and the cylindirical shell formed by the steel. This effect reduces the overall sensitivity of the detector.
The LXe's contribution to the sensitivity is proportional to the intersection length between the LOR and the cylindrical shell formed by the LXe. This effect increases the overall sensitivity of the detector.
These two shells are contiguous, so the geometries' details in the two cases are very similar. Thus these two effects are similar but act in opposite directions.
A quick back-of-the-envelope estimate suggests that the LXe contribution to the sensitivity is very roughly twice the steel's contribution to the attenuation: LXe ≈ -2Fe. So a rough estimate of the combined effect (taking attenuation to be positive) is: extra attenuation = Fe - LXe ≈ -Fe.
Including a prototype implementation of the steel's attenuation in the sensitivity image, exacerbates the St Andrew's cross. Applying the aforementioned trick and subtracting the steel's contribution to get a rough estimate of the total effect of the steel and the LXe, ELIMINATES THE ST. ANDREW'S CROSS ALMOST COMPLETELY.
This has no appreciable effect on the CRCs or SNRs, but it does improve the background variabilities. One would expect the bgvars to improve, as removing the St. Andrew's cross makes the background much more uniform.
Attenuation correction currently takes into account only the material in the FOV. The steel layer in front of the LXe contributes significantly to the attenuation, and is, thus far ignored.
Take this layer into account in scatter correction, and see if it mitigates the St. Andrew's cross artefact.